Wikipedia talk:Edit filter/Archive 4

Special:AbuseFilter/18

Can something be done for addition (accidental or not) of a blank table, such as this:

header 1	header 2	header 3
row 1, cell 1	row 1, cell 2	row 1, cell 3
row 2, cell 1	row 2, cell 2	row 2, cell 3

I have spent the last hour going through articles that had these in them, and I still have thousands of search results to go. Could something possibly be added to Special:AbuseFilter/18 to catch the string "! header 1", "! header 2" or "! header 3", since these are unlikely to appear in constructive edits. Not sure this should be a "disallow" case, but a tag would be very helpful.-RunningOnBrains^(talk) 20:41, 22 July 2009 (UTC)

I added "! header 1" to the filter. Ruslik_Zero 13:41, 23 July 2009 (UTC)

I changed "! header 1" to "| row 1, cell 1." Header 1 might (just might) have a plausible use, while I could not think of one for row 1, cell 1, IMHO. -- King of ♥ ♦ ♣ ♠ 05:40, 24 July 2009 (UTC)

Tags in contributions and histories and tag filter input box

I don't think it's necessary to show tags in histories and contributions. It can be problematic with false positives, and the filter log can be used for verifications, tags are really needed in recent changes, and for other functions we may have in the future, but they are no major needs to have them displayed there. There is the filter log linked in the contributions, and in histories, we could add a "edit filter log" link next to the "View logs for this page" link. Tags can also be applied to logs, see here for examples, there too, it's probably unneeded. Now the tag filter input box, do we need this ? I'm not sure it's really used, or even usable. Cenarium (talk) 00:49, 11 July 2009 (UTC)

*yawn* why are you coming up with every conceivable solution to the problem of false positives except FIXING THE FILTERS THAT CAUSE THEM. Gurch (talk) 22:47, 12 July 2009 (UTC)

I fixed or disabled several filters, removed inappropriate taggings like the Mikael Jackson and unflagged bot filters, removed warn or disallow actions for filters that ought not to have those, e.g when I removed disallow for filter 18. But this is largely irrelevant; this is not so much about false positives, but usability. Cheers, Cenarium (talk) 14:59, 24 July 2009 (UTC)

Filter 52 - Edit summary vandalism II

The activity on this filter is dwindling down to only about 20 hits a month, and it's also prone to false positives. Is it worth the 7.63ms it takes to run? (A huge filter indeed!) -- King of ♥ ♦ ♣ ♠ 04:59, 21 July 2009 (UTC)

Any filter that takes more than 5 ms should be disabled, IMO - if an edit takes too long to process it could time out and just return a BSOD. There ought to be some way to optimize that, although I can't check right now as I'm on the wrong account. Hersfold non-admin^(t/a/c) 17:46, 22 July 2009 (UTC)

I've edited it and it's under 6 ms now, so I've reenabled. This may not get many hits per month, but it is a very important filter for dealing with a particular set of long-term vandalism. Any further ideas to reduce the runtime would be greatly appreciated. NawlinWiki (talk) 14:32, 24 July 2009 (UTC)

Special:AbuseFilter/212

What is the point of this filter? It's generating filter log entries for perfectly appropriate actions. Experienced users tend to edit the final section to add new sections as well. It's annoying, but doesn't warrant an edit filter. –xeno^talk 17:41, 28 July 2009 (UTC)

Experienced users are exempted by the !autoconfirmed. Also, I recently fixed this filter to remove the false positives; don't look at the ones before July 27. The point of the filter is to make sure new users add headers to their posts, either by clicking "new section" or adding "==" around the header. Yes, experienced users do edit the final section, but if new users do that along with "==", it's not going to get flagged. But what about replying to another post? That's an unavoidable false positive - but wait! It's not a false positive after all, since they're supposed to put colons before it to indent. (I know, no need to hassle over a colon, but the filter just conveniently flags those mistakes as well.) The only real false positive I've seen since the correction was users editing a subpage of their own talk (e.g. [1]). -- King of ♥ ♦ ♣ ♠ 22:18, 28 July 2009 (UTC)

I think this is an unnecessary drain and (if set to warn and/or disallow) a barrier to new users getting help. It's a minor annoyance and not something we should engage the EF on - imo. –xeno^talk 22:29, 28 July 2009 (UTC)

Then the warning should be made as friendly as possible. (Of course it's not going to be disallowed.) I think it would help both the new users (learn how to post on talk pages) and established users (not have to get confused because the post was not under a header). They're going to have to learn anyways, so why not earlier? See Special:AbuseFilter/167; isn't that "a barrier to new users" creating articles? -- King of ♥ ♦ ♣ ♠ 23:03, 28 July 2009 (UTC)

Well I guess we'll just have to wait for others to weigh in. I'm not a fan of this filter at all. Just as an aside, please do try to ensure your titles are a little more tactful, especially for a filter like this targeting good faith edits (I'm talking about the original title "New user editing user talk page", not the newer, better one). –xeno^talk 23:06, 28 July 2009 (UTC)

Seems odd that KoH disables a filter which is designed for a single article and takes barely a millisecond, but then creates this one using more than twice the CPU time for edits that cause no damage at all. I'm baffled. Wknight94 ^talk 03:35, 29 July 2009 (UTC)

Yes, I thought it was a little on the expensive side. –xeno^talk 03:44, 29 July 2009 (UTC)

But 212 gets over 20 hits a day. It's the hit/runtime ratio that counts. -- King of ♥ ♦ ♣ ♠ 04:42, 29 July 2009 (UTC)

No, it's useful hits/runtime ratio that counts. 212 is catching stuff no one cares about, i.e. 0 useful hit ratio. Wknight94 ^talk 14:04, 29 July 2009 (UTC)

(unindented) What about Special:AbuseFilter/104? Placing {{helpme}} signifies that you want help, no matter where you put it. It's just that if you put it in the main namespace, someone will have to clean up after you and find you in the history to know who put it there. (Isn't that what we have to do if someone posts on your talk page, neither adding a header nor signing?) And Special:AbuseFilter/211 is for their own good; 212 is (in addition to making it more convenient for us) also for their own good. To say that it is useless is a bit of an overstatement. Let's say that 153 gives you a small gold coin every month, and 212 gives you a dollar every hour. Which would you rather have? -- King of ♥ ♦ ♣ ♠ 16:11, 29 July 2009 (UTC)

104 prevents garbage from going in the mainspace. 211 saves people getting spammed. 212 ... what does it do? Have you been following up the logged reports and teaching users how to use talk pages? They'll learn to use talk pages eventually, there's no need for a an expensive filter. It should be disabled. –xeno^talk 16:14, 29 July 2009 (UTC)

Let's go back to the example of 167. What harm does that do? -- King of ♥ ♦ ♣ ♠ 16:24, 29 July 2009 (UTC)

According to the filter, the need to "brute forc[e] through thousands of submissions to locate [the malformed request]" . –xeno^talk 16:34, 29 July 2009 (UTC)

I also don't see much use in this filter. The edit filter really shouldn't be used to teach newbies how to use talk pages. --Conti|✉ 16:51, 29 July 2009 (UTC)

Seeing that there is no support for this EF, I've disabled it. Question: would it be beneficial to modify MediaWiki:Talkpagetext to teach them to use talk pages? -- King of ♥ ♦ ♣ ♠ 17:03, 29 July 2009 (UTC)

The first link there leads to an explanation of talk pages and the use of the same. I assume most newbies don't read stuff like this and will just do what they're planning on anyway. –xeno^talk 17:48, 29 July 2009 (UTC)

BTW, go ahead and disable Special:AbuseFilter/104 too for that matter. That looks almost as useless. And is it really taking over 7ms for that?! Unreal what my little 1ms filter is being shut off in favor of... Wknight94 ^talk 18:00, 29 July 2009 (UTC)

 Done. Was actually 9.1ms at the time of disabling. {{helpme}} templates, by their very nature, will quickly draw users to their erroneous usage. No need for a filter. –xeno^talk 13:40, 30 July 2009 (UTC)

Retrieving new wikitext for a disallowed edit from the filter log

How can we do that ? Simple copy/paste doesn't preserve spaces (and when you get interminable tables...). See here for such a request. Cenarium (talk) 04:09, 30 July 2009 (UTC)

WP:FFD and Image_name.ext

Think we could have a filter to detect the placing of FFD templates with Image_name.ext as their filename on the FFD page to trigger a warning message? ViperSnake151  Talk  22:04, 1 August 2009 (UTC)

Request for "edit filter managers" permission

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section.

Jakew (talk · contribs · deleted contribs · logs · filter log · block user · block log)

I'd like to request the "edit filter managers" permission. As can be seen from User talk:Avraham#Abuse filter (and my log), this was previously assigned to me for a few hours. The permission was removed citing need for discussion; hence this thread. It was originally granted in response to my volunteering to roll up my sleeves and learn how to create a filter rule, something which interests me and I believe would be helpful to the project. Thanks to User:Avraham, my idea became what is now rule #216.

I'm happy to answer any questions. As a brief overview, I've been editing Wikipedia since 2004. I'm trained as a computer software engineer, and have an engineer's sense of caution. I understand that modifying an edit filter is a serious undertaking, and if granted this permission I'm likely to err on the side of caution. Thanks, Jakew (talk) 16:34, 2 August 2009 (UTC)

Can someone please explain rule #216? DanBlackham (talk) 15:30, 3 August 2009 (UTC)

Sure. There has been a particular form of anon vandalism for a long time now. Follow the link brought in the note section for an example. Right now, #216 is set up to log instances of that particular form of vandalism, which currently is met by immediate reversion. If the vandalism remains semi-regular, and the filter catches it properly, the idea is to prevent that particular kind of edit. -- Avi (talk) 15:37, 3 August 2009 (UTC)

I certainly see no problem in granting it to Jake. I guess this is a stupid question and in the wrong place, but Jake, have you thought of running for adminship? You only RFA attempt was two and a half years ago and at first blush you candidacy does seem possible. MBisanz ^talk 16:55, 3 August 2009 (UTC)

Thanks, MBisanz. I don't think it's a stupid question. To answer briefly, I'm not quite ready to request adminship again just yet, as I think some of the concerns that were raised in my first RfA, namely the need for a broader range of edits including those in the WP namespace, are still valid. Jakew (talk) 18:58, 3 August 2009 (UTC)

Okey, fair enough. Assuming no one else comments in the next couple of hours, I'll assign you the userright. MBisanz ^talk 19:00, 3 August 2009 (UTC)

I originally assigned the userright, incorrectly in hindsight as I was in ignorant of the "Cobi" decision, thus I would support the addition of the right as well. -- Avi (talk) 19:58, 3 August 2009 (UTC)

Ok,  Done MBisanz ^talk 01:14, 4 August 2009 (UTC)

Sounds good! Prodego ^talk 01:26, 4 August 2009 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Question

In the Special:AbuseFilter page, the consqeuence column has the actions taken by the filter, such as warn, tag, and disallow. But I'm uncertain about what happens for the action "Block autopromote" and "Throttle". Can anyone explain about this? Thanks, -- 科学高爾夫迷(讨论|投稿) 02:40, 3 August 2009 (UTC)

I'm not too sure about "throttle", but "block autopromote" stops new users from getting the "autoconfirmed" status which is normally granted after 4 days and 10 edits. →javért ^stargaze 04:47, 3 August 2009 (UTC)

Throttle is for throttling certain actions to X per Y. -- Cobi^(t|c|b) 04:50, 3 August 2009 (UTC)

Okay, so what does "throttling" mean without using the word "throttle?" --68.127.233.138 (talk) 05:44, 3 August 2009 (UTC)

I believe it is limiting the number of performances of a given action. For example, only 1 edit in the next 60 seconds. -- Avi (talk) 05:50, 3 August 2009 (UTC)

Lets say there is a vandal moving pages to names containing 'rock'. Well we can't block all moves to page titles containing 'rock'. But, we could use throttle such that anyone who moves more than 2 pages to titles containing 'rock' per 5 minutes triggers a filter. Then we could set that filter to disallow the move. Prodego ^talk 05:52, 3 August 2009 (UTC)

Nevermind, it's a limit not an action. That's what happens when you try and answer questions at 2AM :} -- Avi (talk) 05:55, 3 August 2009 (UTC)

What I described above is the throttle action. Throttle in conjunction with disallow will create a true throttle, limiting the number of actions that can be done per some period of time. This would be similar to the built in ratelimit. Prodego ^talk 05:57, 3 August 2009 (UTC)

So, "throttle" means to limit the number of times someone can do something. Your 2AM edit was fine, Avi. --68.127.233.138 (talk) 02:09, 4 August 2009 (UTC)

Diffs?

Is there a way to get actual "regular" diffs for filtered actions (a filter that logs but does not prevent an edit). The "details" link gives a kind of diff, but I'm not sure how to get a regular diff of the edit except to go to the page and get it manually. Thatcher 13:08, 3 August 2009 (UTC)

No, you have to go to the page (easiest way is via the contribs link). Prodego ^talk 16:37, 3 August 2009 (UTC)

Lamest catch ever

OMG, you're kidding? There's a filter that tags users that put an ellipsis on user talk pages! Edit filter 135 might be the lamest thing I've found on en.wiki to date.

It probably started by tagging everyone's signatures.

--69.226.103.13 (talk) 21:05, 30 July 2009 (UTC)

You've got one period too many in your ellipsis... –xeno^talk 21:07, 30 July 2009 (UTC)

No, it's an ellipsis at the end of a terminal sentence, 4 is the correct number of dots for such an ellipsis. --69.226.103.13 (talk) 06:55, 31 July 2009 (UTC)

It wasn't the ellipsis, it was the </big> </big> </big> </big> </big> </big></big></big> used in the previous paragraph by an earlier poster. Unfortunately, the edit filter mechanism doesn't let us isolate exactly what the changes were, it includes the paragraph before and after as well - just like a WP:DIFF I imagine. As in other reports, this was just an informative tag that assists in catching a lot of vandalism. Non-vandals can safely ignore. Wknight94 ^talk 09:09, 31 July 2009 (UTC)

But, non vandals will continue to be so tagged. I'm tired of being tagged as a vandal and being told to ignore it and get over it. And, it's a permanent record on this account, but, there's no note saying it wasn't vandalism. This is totally pointless. You don't catch vandals by irritating good editors. Or if you do, then anybody can edit is a lie. --68.127.233.138 (talk) 04:39, 1 August 2009 (UTC)

You know, I'm reasonably sure I've been tagged a few times, and I'm an administrator. Seriously, it does not matter. J.delanoy^gabs_adds 04:42, 1 August 2009 (UTC)

Okay, maybe it doesn't matter to you. Is it necessary to tell me what matters to me? I don't tell you what matters to you. --68.127.233.138 (talk) 04:46, 1 August 2009 (UTC)

What exactly do you hope to accomplish by posting every time you get tagged wrongly? We are not going to disable it, if that's what you are aiming at. J.delanoy^gabs_adds 04:49, 1 August 2009 (UTC)

I hope to make others see how lame and pointless it is. And how pointless to have a permanent record of programming incompetence. Because that's what it really is. Not an edit filter. An edit filter tags the edit, not a random nearby edit. An edit filter tag gives the correct information, not a permanent records of programming incompetence that assigns a tag to a neighboring, unrelated edit. --68.127.233.138 (talk) 05:49, 1 August 2009 (UTC)

It's not personal. It is just a piece of metadata along side the edit. It is not incriminating or a sign of definite abuse, nor is it something to be treated as such. It is simply a note that a snippet of code matches against the edit. Tags are only different from the internal log in that they are publicly viewable. Again, it's not personal. -- Cobi^(t|c|b) 06:06, 1 August 2009 (UTC)

No, they're designed to catch vandals. That's what they were first called. I'm not a vandal. There's no need to improperly add tags to my edits that imply I've done something that needed to be tagged, particularly when it's not even my edit but the edit before mine that triggered the tagging. If it's nothing, don't tag it. If it's something, don't tag it indiscriminately. --68.127.233.138 (talk) 06:18, 1 August 2009 (UTC)

Since the EF has been on there have been no Hagger edits, which means my watchlist isn't filled with redlinks to stupid vandalised pagemoves, if the price for that is that you are tagged for a minor edit (which can be seen by anyone who checks is nothing, and not vandalism) then I don't care if it tags every single edit on Wikipedia, because unless you have been a registered user and had to wake up each morning to fifty Hagger page moves then the current situation of a couple of innocent tags is worth it. Now stop complaining and get back to editing. Darrenhusted (talk) 11:50, 1 August 2009 (UTC)

You sure that's why and how the Hagger page moves were stopped? I have been a registered user. The Hagger page moves stopped appearing in edit histories before these edit filter tags started, since I quit my registered name a while ago-assumed puberty had hit. So, are they related? Where is that mission statement? Don't see it anywhere. Again, if that's the purpose, to stop the Hagger page moves, then stop tagging me. And I'll edit if and when I want to, and voice concerns over problem areas as I see fit. --68.127.233.138 (talk) 02:38, 2 August 2009 (UTC)

Yes the Hagger page moves were stopped by the Edit filter. Darrenhusted (talk) 11:34, 2 August 2009 (UTC)

And, after they were stopped, then the edit filters continued because, after all, there are so many more Hagger moves? To honor the Hagger vandal? I'm still pretty sure it's due to puberty, or mommy needing a little extra help around the house. --68.127.233.138 (talk) 05:35, 3 August 2009 (UTC)

Yes, we get it, Hagger was a pre-pubescent boy, problem was that contantly having to delete and undo those edits took up a lot of man hours, so the EF stops them, and for that we are grateful, apart from you, who seems to have a problem which would only be solved by disabling something the community voted to enable. Darrenhusted (talk) 13:47, 3 August 2009 (UTC)

Yes, of course, the voice of dissent is evil, or at least lacking in gratitude, a quality all proper humans should have... The usual disambiguation into nothingness as a voice of concern is put down. En.wiki had lofty goals that included a community spirit. One vandal changes all that, and turns editors onto personal attacks to defend the change. How sad. --68.127.233.138 (talk) 02:07, 4 August 2009 (UTC)

Did you ever consider that if you are the only person who thinks something is a problem, then maybe it is not really a problem? Because I don't see anyone other than you "voicing concerns" about this. J.delanoy^gabs_adds 03:15, 2 August 2009 (UTC)

Why, because you're not reading other writers complaints? Because you're not looking for them? Because you want to hassle me about my complaints but not bother with other posters who continue to comment about how poorly designed and implemented the filters are? I'll go ahead and keep pointing them out as long as they are inaccurate, poorly designed, and don't do what they were intended to do in addition to inaccurately doing what they intended to do.

Or maybe, like the filters, you aren't identifying the correct edits? --68.127.233.138 (talk) 05:35, 3 August 2009 (UTC)

Everyone gets tagged, but if the edits aren't vandalism then editors move on, something you seem to have a problem doing. There may have been a few complaints but for the amount of vandalism prevented a few false positives are worth it. Darrenhusted (talk) 13:47, 3 August 2009 (UTC)

You're here. And, now that we've moved into discussing the person (my "problem moving on"), not the issue, maybe we should continue in like vein about other personal issues?

It's interesting how many personal attacks my dislike of the poorly designed edit filters gather. The surprise? No better programming of the filters to do what they were designed to do, no supporting links about their designs, no supporting links about their success. And the effort expended to justify the poorly designed filters never extends to fixing the programming. Just personal attacks, eventually. Sure, personally attacking me for pointing out what is wrong with the filtes is easier than fixing them. Programming well is a difficult skill, not randomly available on en.wiki.

One way to move on in cyber space is to stop. It's often passed over for the opportunity to accuse someone else of failing to move on. --68.127.233.138 (talk) 01:59, 4 August 2009 (UTC)

Except that in the last four days you have done nothing other than comment on bots and here, you haven't actually edited any articles. "poorly designed filters"? This would be true if the EF were blocking users incorrectly or disrupting those who are making a contribution, but most of the time if you get caught in a filter (repeating characters, move vandalism, section blanking being the most common) then the truth is you are probably a vandal, but it is still left to editors to decide if the edit was vandalism. So as you are a lone voice complaining about something editors feel is being handled fine by the EF, you talk about how "En.wiki had lofty goals that included a community spirit.", and we do, for those who make a contribution, so far you have not. This is not a personal attack, simply a suggestion: try editing some articles. Darrenhusted (talk) 11:56, 4 August 2009 (UTC)

Edit warring enforcement

I think it would be ideal if an edit filter could be created for every article. That way, not every edit will have to go through the entire sequence of filters just for one instance. In the meantime, however, I believe that we could have certain "temporary" filters (e.g. reserve 221-230, and have them normally set on disabled) designed to block certain users from editing certain pages for a certain duration. The syntax would be like:

(user_name = "King of Hearts") &
(article_text = "Main Page") &
(timestamp < 1249362000)

set to disallow

Because blocks are intended to be preventative, not punitive, this cures the problems with blocking (prevents the user from editing only the disputed page) and page protection (innocent users aren't affected). How does this idea sound? (If you believe this wastes too much resources, then also comment on whether you would support this idea if article-specific filters were created.) -- King of ♥ ♦ ♣ ♠ 23:56, 3 August 2009 (UTC)

You mean filters that would enforce Wikipedia:Editing restrictions? I don't really think that's what the edit filter was created for. The current system of blocking users who break/ignore their restrictions seems to work quite fine. --Conti|✉ 13:28, 4 August 2009 (UTC)

Filter 139

Filter 139 prevents fixed position vandalism. It should generally disallow for userspace because of userboxes and other transcluded content, but users should be allowed to use this in their own userspace. Any idea how to allow this ? article_text shouldn't simply contain user_name, otherwise it could be circumvented using user names contained in the targeted username. So either it should be the userpage or "User:user_name/" should be contained in "article_prefixedtext", but the edit filter doesn't seem to handle. Cenarium (talk) 15:43, 7 August 2009 (UTC)

 Done. Just a little bit inefficient, but I couldn't think of a better way to group the two checks into one. -- King of ♥ ♦ ♣ ♠ 18:28, 7 August 2009 (UTC)

Special:AbuseFilter/131

I am probably doing something wrong, but I can not make the filter work on Rorschach test article. The filter does catch anything. Ruslik_Zero 12:21, 7 August 2009 (UTC)

(regex noob, but...) Shouldn't [01][0-9] be ([01]|[0-9]) ? –xeno^talk 12:32, 7 August 2009 (UTC)

I'm no regex expert, but I don't think so. Since the set (0,1,2,...9) includes the set (0,1), ([01]|[0-9]) is equivalent to [0-9]. [01][0-9] should catch all two-digit sequences from "00" to "19". I'd guess that's probably what's wanted. I've no idea as to the problem, unfortunately... Jakew (talk) 12:38, 7 August 2009 (UTC)

Yes, I meant exactly this. Ruslik_Zero 12:41, 7 August 2009 (UTC)

kk. As I said... regex n00b... just throwing it out there. –xeno^talk 12:42, 7 August 2009 (UTC)

Could admins please get into the habit of noting where the consensus was formed for such an action as preventing users from removing an image? :) --Conti|✉ 13:21, 7 August 2009 (UTC)

On Talk:Rorschach_test. Ruslik_Zero 15:37, 7 August 2009 (UTC)

That page is currently 348 kilobytes long. Could you be a bit more specific? Anyhow, my point was that such a link should be provided when the filter is modified/created, as a courtesy to those that have not followed the issue. --Conti|✉ 15:45, 7 August 2009 (UTC)

Banner at the top of the article. Ruslik_Zero 16:56, 7 August 2009 (UTC)

Which of the 10 banners? ;) Nah, found it now, thanks. --Conti|✉ 17:03, 7 August 2009 (UTC)

The problem was in the following expression

(article_text == "Muhammad" | 
 article_text == "Temple garment" | 
 article_text == "Endowment (Latter Day Saints)" | 
 article_text == "Rorschach test"
)

As I discovered, after the word "test" and before the " mark there was an invisible symbol. After I typed the title manually without copy-pasting the filter began to work. Ruslik_Zero 18:52, 7 August 2009 (UTC)

Not that I am endorsing this filter, but the overlay image can still be removed, FWIW [2]. –xeno^talk 19:03, 7 August 2009 (UTC)

Not now. Ruslik_Zero 19:48, 7 August 2009 (UTC)

The titling seems inefficient, why not (contains_any(article_text,"Muhammad","Temple garment","Endowment (Latter Day Saints)","Rorschach test") Would using article_articleid be faster?(There's no documentation={)Smallman12q (talk) 01:50, 10 August 2009 (UTC)

RFC in devolping criteria

Rfc for devolping criteria for granting permissions at Wikipedia:Village_pump_(policy)#Criteria_for_Abusefilter.2Feditfilter_permissions.Smallman12q (talk) 21:43, 10 August 2009 (UTC)

Efficiency

Am I correct in understanding that when targeting one or a small group of articles, putting the check for the article ID's first makes for the most efficient use of the filter? -- Avi (talk) 14:39, 2 August 2009 (UTC)

Yup. I believe the more "rare" it is, the further up it should go.Smallman12q (talk) 20:26, 11 August 2009 (UTC)

Requested Edit Filters

Hi there all, I was looking at edit filters today and noticed something on the requested page, so I figured I would give it a shot. Could the filter at Special:AbuseFilter/103 be replaced with this?

!("sysop" in USER_GROUPS) &
(article_namespace == 0) &
(("uncyclopedia" in lcase(added_lines)) &
("oscar wilde" in lcase(added_lines)) &
(("uncyclopedia" in lcase(old_wikitext)) |
("oscar wilde" in lcase(old_wikitext)))


I changed the user_group to sysop because I figured that there isn't really a reason for a rollbacker, autoconfirmed user, etc to add the words to an article that doesn't already contain them. Any tips, suggestions, or comments are welcome. :) Thanks for reading, MacMed^talk_stalk 05:55, 7 August 2009 (UTC)

What is "oscar wilde"? Ruslik_Zero 11:19, 7 August 2009 (UTC)

The Uncyclopedia site commonly uses quotes (real and imaginary) from Oscar Wilde. Looking at it, I changes the OR operator (|) between the "uncyclopedia" condition and the "oscar wilde" condition to an AND operator, since additions of Oscar Wilde on his own would create too many false positives. So perhaps a new rule for this, since not all Uncyclopedia additions will mention Oscar Wilde. Regards, MacMed^talk_stalk 14:59, 7 August 2009 (UTC)

I believe that the above code would return a hit any time a section containing the word "Oscar Wilde" is edited, even if it is not removed, as it is returned in the added_lines section. Also, why is the filter being triggered if "oscar wilde" is already in the old wikitext? It should be when the words are not already there (missing a ! it seems). -- Avi (talk) 15:10, 7 August 2009 (UTC)

Actually, would this be better?

!("autoconfirmed" in user_groups) &
article_namespace = 0 &
count("uncyclopedia",lcase(added_lines)) > count("uncyclopedia",lcase(removed_lines))

-- Avi (talk) 15:16, 7 August 2009 (UTC)

But what if somebody is adding the word Uncyclopedia to a section of the Uncyclopedia article? The lines they take out may not neccessarily contain "uncyclopedia", but they are adding it, so the filter would still trip with a false positive, would it not? And by the way, yes I was missing the ! in the old code. Regards, MacMed^talk_stalk 15:27, 7 August 2009 (UTC)

If that is a problem solely in that article, it's simple enough to add a trap to ignore "Uncyclopedia". Also, if what you really want is not to trip the filter any time the word "uncyclopedia" already exists in the article, I'd put "!("uncyclopedia" in lcase(old_wikitext))" first, to take advantage of the short-circuit evaluation. Since EVERY time it exists you want the filter to fail, but /not/ every time it is added you want it to succeed. -- Avi (talk) 15:33, 7 August 2009 (UTC)

So something like this:

!("autoconfirmed" in USER_GROUPS) &
!("uncyclopedia" in lcase(old_wikitext)) &
(article_namespace = 0) &
count("uncyclopedia",lcase(added_lines)) > count("uncyclopedia",lcase(removed_lines))

might work? MacMed^talk_stalk 15:47, 7 August 2009 (UTC)

KoH, DF, and Prodego are wizards at optimization, but I think this would be a bit better:

!("autoconfirmed" in USER_GROUPS) &
  ((article_namespace = 0) & 
   ( !("uncyclopedia" in lcase(old_wikitext)) & (count("uncyclopedia",lcase(added_lines)) > count("uncyclopedia",lcase(removed_lines))
    )
   )
  )

Although the additional parenthases are considered an extra heck, it also groups together the operations for short-circuiting purposes. This way, while a true edit would end up with something like 9 checks on the edit, at any point it fails except the very end, it short circuits, so if the user is autoconfirmed, there is only one check; a non-autoconfirmed IP in any space but article dies after only 3 checks, and so on. "In", IIRC, is a more resource intensive check, so I think it should go later than the first two. I'd like someone more proficient than I to comment, though :) -- Avi (talk) 16:12, 7 August 2009 (UTC)

'oscar wilde' and 'uncyclopedia' both need to be detected at the same time.--Otterathome (talk) 16:26, 7 August 2009 (UTC)

So what about:

!("autoconfirmed" in USER_GROUPS) &
((article_namespace = 0) & 
( !("uncyclopedia" in lcase(old_wikitext)) & (count("uncyclopedia",lcase(added_lines)) > count("uncyclopedia",lcase(removed_lines))
& ( !("oscar wilde" in lcase(old_wikitext)) & (count("oscar wilde",lcase(added_line)) > count("oscar wilde",lcase(removed_lines))
)
)
)

Could that work? MacMed^talk_stalk 17:30, 7 August 2009 (UTC)

I think this would be better, but I'd like to hear someone more experienced than I on its efficiency:

!("autoconfirmed" in USER_GROUPS) &
 ((article_namespace = 0) &
  (!("uncyclopedia" in lcase(old_wikitext)) & (count("uncyclopedia",lcase(added_lines)) > count("uncyclopedia",lcase(removed_lines)))) &
   (!("oscar wilde" in lcase(old_wikitext)) & (count("oscar wilde",lcase(added_lines)) > count("oscar wilde",lcase(removed_lines))))
  )

-- Avi (talk) 17:37, 7 August 2009 (UTC)

It is not supose to be a replacement for Special:AbuseFilter/103 as that filter is working great but has tripped some inappropriate, though good faith edits so cannot be set to Disallow. This is an additional filter that detects and disallows edits that cannot possibly be done in good faith and cannot possibly be constructive edits.--Otterathome (talk) 17:55, 7 August 2009 (UTC)

Perhaps we could do a regex: {{Q|.*?|Oscar Wilde}}. This would be an OR condition, since {{Q}} is not used for quotations on Wikipedia. -- King of ♥ ♦ ♣ ♠ 17:59, 7 August 2009 (UTC)

How would you escape the text in the regex so that the | in the template isn't seen as an OR by the regex? MacMed^talk_stalk 22:54, 7 August 2009 (UTC)

Oops, I meant {{Q\|.*?\|Oscar Wilde}}. -- King of ♥ ♦ ♣ ♠ 21:40, 8 August 2009 (UTC)

If it's that much trouble then there's no need, as Special:AbuseFilter/103 will pick it up, but it only warns the user. Putting it to disallow would be risky.--Otterathome (talk) 09:49, 8 August 2009 (UTC)

There's absolutely no good reason to display a question mark with two parameters, the second of which is "Oscar Wilde." -- King of ♥ ♦ ♣ ♠ 16:48, 10 August 2009 (UTC)

Another vandal edit just happened again that this filter would have disallowed[3].--Otterathome (talk) 21:12, 11 August 2009 (UTC)

Filter 200, or should the EF be engaged to track non-abusive, non-"wrong" edits?

Special:AbuseFilter/200

Why on earth is there a filter looking for removed prods that is tagging edits with "tag: prod removed"? I've triggered this thing a number of times now and to be perfectly honest I can't see that it has any purpose whatsoever. This filter makes it seem as though you are doing something wrong when you remove a prod; both the tag and the filter log. If an editor wants to know if a prod they placed on a particular article has been removed then they need to watchlist said article. --Tothwolf (talk) 16:47, 13 July 2009 (UTC)

• I agree, this filter is a bad idea and is just going to cause more negative feelings about the edit filter. I've disabled it. I'm sure a bot can be written to monitor the prod category without needing a filter to help it along. –xeno^talk 16:52, 13 July 2009 (UTC)
  • See Wikipedia:Edit filter/Requested. SoWhy believes that the filter would make his bot run more efficiently. -- King of ♥ ♦ ♣ ♠ 17:10, 13 July 2009 (UTC)
    • I don't think the EF was designed to be, nor should it be, used as part and parcel of a bots' operation. However, if I am alone in this opinion I don't object to the filter being re-enabled, but I would like to see more opinion on this. At present, people who see entries in their edit filter log don't take kindly to it, no matter how much we tell them that it doesn't necessarily mean it's an abusive edit. –xeno^talk 17:12, 13 July 2009 (UTC)
      • I think the filter is useful not only for a bot (the bot in question was previously requested by another user and uses a different mechanism). But I thought we have renamed the abuse filter specifically because not all edits tagged are really abuse and so I thought consensus was by now that filters can be used for any kind of edits, not only abusive ones. I think the filter is useful to highlight the removal of prods for people who watchlist those articles as tags show up much better than simple edits (where the edit summary usually lacks this particular information). It's of course a matter of consensus whether this filter should be enabled but "negative feelings" sounds like a weak argument. The filter serves to make maintenance easier, doesn't it? So such filters cannot be negative per se just because the edits are not abusive. It's similar to Special:AbuseFilter/29 in this regard. Not all edits this filter flags are abusive but it's informative to have. Regards SoWhy 17:26, 13 July 2009 (UTC)
        • Filter 29 tags new users. This one tags all users, so we're going to have a lot of folks generating entries in their EF logs, and a lot more complaints about that. We've renamed the filter yes, but I think that was because that in looking for abuse, it sometimes tags non-abusive edits - not because we should now use it for stuff like this. As I said above, I'm prepared to be wrong on this. –xeno^talk 17:29, 13 July 2009 (UTC)
          • I, too, am prepared to admit to be wrong but I had thought that with the name change and with some filters, consensus changed on this issue. For example, filters 183, 155 and 96 track contributions that are mostly or always good-faith and non-abusive. Regards SoWhy 17:40, 13 July 2009 (UTC)

← They track good faith contributions that are usually wrong, whereas removing prod warnings (unless you're that abusive socky guy) is almost never wrong. –xeno^talk 17:42, 13 July 2009 (UTC)

Well, adding links to YouTube is not per se "wrong", is it? People just use them too often but that does not make the edits themselves wrong or abusive and that's the point I want to make. Yes, removing prod templates is almost always correct but imho still useful to know. Regards SoWhy 17:50, 13 July 2009 (UTC)

They're usually wrong, being right is the exception. Vice-versa for removing prods, so generating a note in someone's EF log every time they do so isn't a good idea (imo). –xeno^talk 17:55, 13 July 2009 (UTC)

The Article alerts bot can track prods without needing a special edit filter in place so I don't see why another bot couldn't also do so. That said, I'm not even sure a bot is needed to notify someone of removed prods, they can just as easily add the article to their watchlist.
As for Youtube links, I'm unfortunately all too familiar with that filter as I got involved with the discussion over that one when it was extended in an attempt to track torrent links (which was later removed as it is unworkable and unnecessary). I added a valid Youtube link awhile back where a software developer was discussing software functionality. Even though that might be the "exception", and even though it is a perfectly acceptable link (actually used as an inline citation), having it show up in the log with my contribs makes it seem like I did something wrong there too. --Tothwolf (talk) 18:10, 13 July 2009 (UTC)

I would like a way to track deprods. If it can't be added back to the edit filter, then is there another way to have a global list of all deprods? It is useful for seeing if articles can be improved after a deprod, if the deprodding editor didn't, for spotting serial deprodding that may be disruptive, and for filtering deprodded articles into AfD discussions if appropriate. Fences&Windows 00:51, 1 August 2009 (UTC)

A bot could be written to do this. –xeno^talk 12:26, 12 August 2009 (UTC)

Createaccount not working

I was able to successfully create Aaron Black third test (talk · contribs) past filter 225, and KoH Guy test (talk · contribs) without getting logged by 159. Looks like the createaccount check is not working somehow? -- King of ♥ ♦ ♣ ♠ 21:51, 12 August 2009 (UTC)

225 is missing parenthesis and 159's regex wouldn't match "KoH Guy test", so the behavior is expected. (I'd be happy to elaborate if you would like, by the way) Prodego ^talk 22:25, 12 August 2009 (UTC)

Requests for permissions

There have now been a few non-admins asking for access. Should we make a subpage at Wikipedia:Requests for permissions and direct them there to make their request? -- King of ♥ ♦ ♣ ♠ 03:30, 7 August 2009 (UTC)

The only problem I see with that is RFPERM may not be watched by as many folks who are familiar with the EF and thus might not be able to vet candidates as well. (Though I notice we haven't really been doing much vetting as it is...no trick regex questions? no filter test case questions? =) –xeno^talk 03:47, 7 August 2009 (UTC)

What's regex (/me ducks) -- Avi (talk) 06:03, 7 August 2009 (UTC)

Well it's not as if those of you with the permission have much grasp of regexes either (and that list hasn't even been updated since April) Gurch (talk) 23:47, 13 August 2009 (UTC)

I don't think you need to as so few users are going to get edit filter permission anyway because it's too easy to FUBAR up the site with it.--Otterathome (talk) 06:37, 7 August 2009 (UTC)

I still think this right is rare enough that granting it like we do IPBE (informal discussion at the subject matter area) is preferable to the WP:PERM page. Once the volume gets larger here, I would suggest moving it. MBisanz ^talk 15:05, 7 August 2009 (UTC)

Removing copyright tags

Who then was a gentleman? asked at a recent AN thread whether there was an abuse filter to note removal of copyvio tags. So far as I know, there isn't, and it seems like this could be very useful, as the premature removal of these sometimes restores copyrighted text to publication. I've had nothing to do with edit filters and so am unsure of their capabilities (looks complex :)). Ideally, it should flag for human review, since articles tagged with {{copyvio}} aren't reviewed for seven + one days after tagging. These tags are supposed to be removed only by administrators; however, there are many cases where I've seen that non-admins have reviewed and appropriately removed the tags where they were unwarranted. OTOH, most of the time they're removed improperly. There's probably not much call for them to be removed by new accounts or unregistered users, given the high incidence of abuse.

Is this something an edit filter could help with? Does anybody have any input on what such a filter might do? --Moonriddengirl ^(talk) 12:05, 12 August 2009 (UTC)

 Done - Special:AbuseFilter/224. It checks for non-autoconfirmed users in the main namespace making edits such that the number of copyvio tags in the new revision is less than the number of copyvio tags in the old revision. -- King of ♥ ♦ ♣ ♠ 21:07, 12 August 2009 (UTC)

Thank you very much. :) --Moonriddengirl ^(talk) 11:31, 14 August 2009 (UTC)

Hidding the log

For some filters such as Special:AbuseFilter/76 and Special:AbuseFilter/211, the logs should be hidden, otherwise the filter makes it easier for bots. (The logs of private filters are not hidden). By hidden, I mean that you require abusefilter/sysop privileges to see them. Perhaps an enhancement bug should be filed?Smallman12q (talk) 15:20, 9 August 2009 (UTC)

I don't see why we should remove the ability for anyone to view the history of public filters. FunPika 16:00, 9 August 2009 (UTC)

Well, in some case, the filters are being used to prevent people from adding personal info such as emails. By allowing the log to be public, these peoples' emails are still being posted...defeating the purpose of the filter.Smallman12q (talk) 21:37, 9 August 2009 (UTC)

That does make sense; the logs provide easy prey for spammers. -- King of ♥ ♦ ♣ ♠ 17:43, 9 August 2009 (UTC)

Do you have specific evidence? Ruslik_Zero 19:05, 9 August 2009 (UTC)

Evidence? Please elaborate.Smallman12q (talk) 21:35, 9 August 2009 (UTC)

I think there was a suggest earlier about hiding the logs from public view after x number of days for a number of reasons... Spambots harvesting the email addresses viewable in examine would certainly qualify as a reason. The ability to restrict log display to a limited usergroup right away on certain filters also has value, as with the above example. –xeno^talk 22:21, 9 August 2009 (UTC)

It is my opinion that the community interest in overseeing filter performance (false positive rate and so on) outweighs any benefits that hiding the log might give. — Werdna • talk 09:56, 12 August 2009 (UTC)

This isn't for filter performance, this is for privacy. Not hiding the log in some cases makes the filter completely ineffective;even harmful eg: a filter was designed to detect and prevent emails from being posted, but since a log is kept, those emails can be easily harvested by a spam bot.Smallman12q (talk) 15:13, 12 August 2009 (UTC)

But we all know perfectly well that you would all go and hide every private filter log from public view anyway, and then the rest of us would be completely screwed, as we wouldn't be able to go through the logs and find which edits were prevented for no good reason and try to guess why based on their content, so you lot would pretty much have free rein to prevent anything you like without any sort of accountability. No thanks. Gurch (talk) 23:39, 13 August 2009 (UTC)

Could the log have a bit that oversight could set, so that the item is not shown in the log (the edit would still test positive in the testing facility, but well). It would not remove the item from the log, and could still be 'shown' as 'oversighted item'. That makes it possible to count it in the false positive rate (like a neutral vote in a vote). --Dirk Beetstra ^{T C} 10:04, 12 August 2009 (UTC)

There's about 1000 active people to check these, just as to check deleted versions. We're all available. Anyone else who wants to work with them can & should apply for filter permission. DGG ( talk ) 21:59, 18 August 2009 (UTC)

Disabling/Removing Filter 1

This seems to be a little ridiculous. The only thing it does is flag a revision in the edit log. If I am not mistaken, this was meant to only be a test. Barista Girl (talk) 02:18, 14 August 2009 (UTC)

You might want to ask Prodego (talk · contribs) about it. –xeno^talk 16:00, 14 August 2009 (UTC)

Arbcom asked me for it. There are several filters that only log. Prodego ^talk 16:02, 14 August 2009 (UTC)

Sometimes it is important to know how often something happens, and these kinds of filters are excellent for that. And yes, ArbCom did ask Prodego for that. -- Avi (talk) 16:04, 14 August 2009 (UTC)

Why is it called "Misc test filter", then? --Conti|✉ 17:24, 14 August 2009 (UTC)

Because it started out as a test filter for the Edit filters (note, it is #1) and now has been used for various purposes by customizing the code. There is no reason to change the name, and for short-term filters, why create a new one that will have to be deleted in a matter of weeks when a customizable one exists? -- Avi (talk) 17:31, 14 August 2009 (UTC)

Well, my understanding of a "test filter" was that it was used to test something for a few minutes or hours or maybe days, but not weeks. If it weren't for this thread, I would've assumed that the filter has long since served its purpose and can be changed/disabled by anyone who needs it again. --Conti|✉ 17:37, 14 August 2009 (UTC)

Well, it is in use now; when ArbCom/Prodego are done with it, I guess it can be disabled until another need pops up. -- Avi (talk) 17:38, 14 August 2009 (UTC)

• I'm curious why Arbcom wants to track people saying fuck or cunt outside articlespace... Is there a link to the request somewhere? –xeno^talk 17:32, 14 August 2009 (UTC)

I saw the request on func-l (and actually started building a filter for it) and can confirm it. As for why, you will have to e-mail func-l or arbcom-l and ask for that information. As a non-Arbcom member I don't think it is my place to explain it; let ArbCom decide if they want to. -- Avi (talk) 17:34, 14 August 2009 (UTC)

I've written them and suggested a public statement be made about this filter. Note it's fairly expensive as well. –xeno^talk 17:46, 14 August 2009 (UTC)

• After some drama over whether or not "established users" should be sanctioned for swearing at other editors, I thought it would be useful to gather some actual facts on how often users swear at each other in anger, and what the outcomes are. I chose the word "fuck" as a proxy for user incivility, fully knowing that it was an incomplete proxy, and an arbitrator asked to add the word "cunt." Jimbo heartily endorsed the research project to gather this information, if that means anything. The filter captures the use of the word "fuck" and "cunt" (as well as derivatives, compounds, etc) when used by editors with more than 500 edits (to screen out simple vandalism). I examine the log periodically and find cases where editors are using the word against other editors in anger (ignoring milder uses such as "Who gives a fuck") and tracking the response, if any. My research is at User:Thatcher/Sandbox4. I planned to collect 50 or 100 events, to have a good sample, but uses of the words in anger are less frequent than I might have imagined, so I will probably stop after 30 days. Thatcher 18:42, 14 August 2009 (UTC)

  Thanks for the explanation. Respectfully, I don't think the Edit filter should be tasked on this. It will only increase the pre-existing apprehension towards it. Can't this type of data-gathering be done with the database dumps, or searches? –xeno^talk 18:52, 14 August 2009 (UTC)

I don't know. Thatcher 19:02, 14 August 2009 (UTC)

As far as "apprehension" is concerned, the signal to noise ratio is so low that this should actually convince people that using the edit filter to enforce civility is not terribly practical. Thatcher 19:11, 14 August 2009 (UTC)

It could be improved a bit, but civility is ultimately something that depends on context, and can not automatically be identified. The filter is named 'Misc test filter' because I use it for miscellaneous tests. It is expensive in terms of run time, not terribly so in terms of conditions. However, the users who are affected by the run time on this filter are really not being affected by any others, so it balances out and isn't a huge problem. Prodego ^talk 21:35, 14 August 2009 (UTC)

I'll echo Thatcher's words above; the purpose of the data collection was simply to see how often the terms were used, and how often they were used in a clearly uncivil way. And it wasn't an "official" Arbcom or Functionaries request, although obviously both groups have a very strong interest in this area, and we talked about it. Having looked at a sample of the posts caught by the edit filter, I might have added a few more to the list than did Thatcher (although mostly because the posts appeared to me to be uncivil in toto,even if the use of "f***" or "c***" or their derivatives weren't specifically uncivil); however, most of the time the terms were being used conversationally, except for the debate about the DYK of yesterday. I suspect that says something too, but not necessarily about civility. I also agree with both Thatcher and Prodego that this experiment has also shown that it isn't words themselves that are uncivil, it is the context in which they are used that determines incivility. I can't imagine an edit filter being helpful in enforcing civility - but now we have the evidence to show that. Risker (talk) 05:26, 15 August 2009 (UTC)

Who ever suggested that? The fact that it was even considered is pretty alarming. –xeno^talk 14:49, 18 August 2009 (UTC)

As far as I know, it wasn't. Prodego ^talk 14:54, 18 August 2009 (UTC)

Constitutionally, it's unclear whether ArbCom has the authority to set up a filter. But the edit filter was really not designed for this kind of things, and this one is extremity expensive. There are other, more efficient, ways to gather this data (e.g. bots). Unless ArbCom makes a formal statement that this edit filter should not be disabled (and it's constitutionally unclear whether they can), this is up to the usual consensus to decide. For my part, I believe this is way too expensive for such a low-priority data-gathering task. Cenarium (talk) 16:53, 15 August 2009 (UTC)

Creating the filter was something I did, not arbcom. Arbcom (or more correctly, the functionaries) may have asked for it, but ultimately I am the one who is responsible for that filter. This one seems to be very expensive. However, the users it is expensive for are affected by almost no other filters, so the actual load is still far under what I would consider acceptable. Prodego ^talk 22:04, 15 August 2009 (UTC)

ArbCom, you are a pain at the best of times, if you want statistics can you not get them in the way as the rest of us have to rather than using your "power" to demand abuse filters? Oh and BTW "fuck cunt" Gurch (talk) 22:01, 16 August 2009 (UTC)

No one demanded anything, I created the filter because it was requested and the request seemed reasonable. Prodego ^talk 00:44, 17 August 2009 (UTC)

• The filter should be disabled. There are less intrusive ways to gather data about potentially uncivil language than generating filter log entries. –xeno^talk 14:48, 18 August 2009 (UTC)

• If there is a better method, then that should certainly be used. What do you have in mind? Prodego ^talk 14:51, 18 August 2009 (UTC)

• WP:AWB has a fairly robust database scanner. –xeno^talk 14:53, 18 August 2009 (UTC)

• That will not give you data on new events, and it will not identify only experienced editors. Prodego ^talk 14:55, 18 August 2009 (UTC)

• The dumps are updated weekly. You could filter it to exclude archives. Someone could comb it to identify experienced users (someone has to filter it anyways, to exclude people linking to Gropecunt Lane and the line).
• In any case, I don't think looking at usages of fck and cnt is a particularly effective way to study civility. Why not just watch WP:WQA? –xeno^talk 14:58, 18 August 2009 (UTC)

• That is a much more tedious task though, we have a perfectly good edit filter that can identify the edits more easily. Either way the same information gets identified. It should be possible to write a bot to work of the dumps, that would search for a word, identify who added it, and then check if that editor meets some criteria. But that would be slow, and a lot of code, whereas the abuse filter is faster and much easier to write. (As for Gropecunt Lane, it isn't too hard to filter that out, I haven't on off chance it would miss something). This filter isn't really about incivility, it is about the use of uncivil terms, and how it is received. Is the usage of these terms commonplace and accepted? Or not? Prodego ^talk 15:05, 18 August 2009 (UTC)

• It's all too Orwellian for my liking. –xeno^talk 15:07, 18 August 2009 (UTC)

(<-)Being that the exact same information can be culled from other sources, as you yourself suggest above, what is so Orewellian? -- Avi (talk) 15:12, 18 August 2009 (UTC)

[edit-conflict]I suppose it is somewhat inevitable. If Wikipedia is big enough no one person can say what is or isn't acceptable to the whole community. Either way the same information is collected, it's just a question of how. Prodego ^talk 15:14, 18 August 2009 (UTC)

@Avi, the fact that we're tagging folks as having potentially ungood thoughts. –xeno^talk 15:17, 18 August 2009 (UTC)

Euphemism that it may be, I see the "Misc test filter" as analogous to the Verbal Morality Statute, fining editors (via log entries) for their usage of foul language. –xeno^talk 15:54, 18 August 2009 (UTC)

Xeno, you seem to be misrepresenting the filter's actions. No one is being tagged by the filter. Thatcher is taking a collection of edits (that may be obtained via a filter, via a data dump, or via 1000 monkeys dutifully copying every word posted on wikipedia) and manually identifying which ones, in his opinion, indicate usage of those terms in a pejorative manner. And yes, telling someone to "F*** off" usually indicates "ungood" thoughts, xeno. -- Avi (talk) 15:57, 18 August 2009 (UTC)

I welcome the Brave New World that awaits the conclusion of this study! Joking aside, I find the truly problematic instances of incivility come from a sharp tongue free of uncouth language. Delivered by those the community has determined to be exempt from CIV, mind you. –xeno^talk 16:00, 18 August 2009 (UTC)

The edits are being logged in a (at present, permanent) filter log. Yes they aren't capital-T "Tag"ged in the filter sense, but they are still being tagged nonetheless. –xeno^talk 18:40, 18 August 2009 (UTC)

I would point out that full history dumps for enwiki aren't available, so a database dump would only find instances when the word was on the current version of the page at the time the dump was made. Mr.Z-man 16:13, 18 August 2009 (UTC)

Request for abusefilter(editfilter) permissions

The following discussion is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

---

Request failed.Smallman12q (talk) 14:31, 20 August 2009 (UTC)

I'd like to make a request for abusefilter(editfilter) permissions. Mainly, I'd like to help in the debugging, creation, implementation, documentation(which is sorely lacking) and optimization of filters. There doesn't appear to be any set criteria for being granted abusefilter permissions(though its mostly admins), but I believe I qualify. I'm willing to answer questions regarding proper syntax and whatnot.Smallman12q (talk) 14:48, 9 August 2009 (UTC)

Edit Filter Manager is becoming quite popular. Ruslik_Zero 19:06, 9 August 2009 (UTC)

I shall be WP:BOLD and ask a question. If this process seems successful, we might be able to systematize it. The following is an example filter that checks for the words "ass" or "asshole." Correct any errors and optimize it. (Note: You do not have to check for donkey-related false positives.)

(added_lines rlike "asshole?") &
!(autoconfirmed in user_groups) &
!(removed_lines rlike "asshole?") &
(article_namespace == 0)

Cheers, King of ♥ ♦ ♣ ♠ 20:46, 9 August 2009 (UTC)

---

Certainly:

(article_namespace == 0)
& !("autoconfirmed" in user_groups)
& (action == "edit")
& (lcase(added_lines) rlike "ass(hole?)"
& !(lcase(removed_lines) rlike "ass(hole?)"

Or for checking to make sure there's no donkey/ass in the title text/before (this one isn't too good)

(article_namespace == 0)
& !("autoconfirmed" in user_groups)
& (action == "edit")
& !(lcase(article_text) rlike "(donkey|ass|anal|butt)")
& !(lcase(old_wikitext) rlike "(ass(hole?)|butt)")
& (lcase(added_lines) rlike "ass(hole?)")
& !(lcase(removed_lines) rlike "ass(hole?)")

I've based it on Special:AbuseFilter/11, Special:AbuseFilter/39, Special:AbuseFilter/46. I hope it works as I don't have access to the debugging tools=P. Smallman12q (talk) 21:35, 9 August 2009 (UTC)

Unfortunately, it seems that you have made two regex errors in asshole?. What this will check for is asshol or asshole. The correct regex is ass(hole)?\b. The parentheses serve the purpose of grouping, while the \b serves the purpose of ending the word. (Otherwise, assist, asset, etc. would get flagged. \b literally means "any non-word character.") I would suggest that you take a look at this handy reference. -- King of ♥ ♦ ♣ ♠ 22:14, 9 August 2009 (UTC)

=(.You are correct. Well I read up on regex and came up with this a[s$][s$](h[o0][l1]e?|)\b. It catches ass pretty well=P. When I ran it through regextester however, it catches a$$hat and a$shat, but not asshat? Well, I did fail your first test, but I still hope you will see my willingness to learn and assist as the determining factor.Smallman12q (talk) 00:48, 10 August 2009 (UTC)

Don't worry. Regex isn't hard. :-) King of ♥ ♦ ♣ ♠ 00:54, 10 August 2009 (UTC)

Question 2. Check for the following:

Brian is a bold text.
Brian is a headline text.
Brianna is a bold text.
Brianna is a headline text.

Do a batch test on my contribs, with "Show changes that do not match the filter" enabled. You should see the ones with edit summary "Smallman12q #[1-4]" positive and "Smallman12q #[5-8]" negative. -- King of ♥ ♦ ♣ ♠ 17:25, 10 August 2009 (UTC)

(My name is not Brian btw.) -- King of ♥ ♦ ♣ ♠ 17:26, 10 August 2009 (UTC)

For this exercise, you will not need autoconfirmed, namespace, etc. stuff, just the regex. Just a comment (for reference): For the other stuff, (action == "edit") is unnecessary, since pretty much all !autoconfirmed changes are edits. -- King of ♥ ♦ ♣ ♠ 17:29, 10 August 2009 (UTC)

I forgot you can't do a batch test without the permission. I guess you'll need to do it manually: http://regexpal.com. -- King of ♥ ♦ ♣ ♠ 18:02, 10 August 2009 (UTC)

Well I found the page history at User:King_of_Hearts/Sandbox/AbuseFilter. I'm a bit confused as to what you want me to check for though? Is this it: Brian(na|) is a (bold|headline) text.\b(I don't think the \b is needed at the end).Smallman12q (talk) 22:00, 10 August 2009 (UTC)

See my talk page for a problem. Yes \b is unnecessary, as no one would append a character immediately after a period. (Even if they did, so what?) -- King of ♥ ♦ ♣ ♠ 22:10, 10 August 2009 (UTC)

*cough*. You are both wrong. .(Period) isn't a literal period. .(Period) matches any single character. So text.\b would match "texta ", "text1 ", etc. Prodego ^talk 22:00, 11 August 2009 (UTC)

*cough cough*. Prodego, you're also wrong. When I say "see my talk page," that's what I mean. (Wrong as in, not the regex, but the evaluation of the problem.) -- King of ♥ ♦ ♣ ♠ 23:37, 11 August 2009 (UTC)

*hack wheeze* Read the problem! Nonsense! That sounds like work! :) (I was basing my comment on "\b is unnecessary, as no one would append a character immediately after a period.") Prodego ^talk 03:09, 12 August 2009 (UTC)

SpecialFilter#414 - IF count_cases(cold symptoms) > 2 DISTRIBUTE COUGH DROPS. -- Avi (talk) 03:12, 12 August 2009 (UTC)

So...when I find out whether I qualify or not...?Smallman12q (talk) 20:24, 11 August 2009 (UTC)

Shouldn't we wait and see the outcome of the Village Pump topic you opened? Regards, MacMed^talk_stalk 21:09, 11 August 2009 (UTC)

Sure, why not. Hopefully there will be more discussion(perhaps it should be added to centralized discussion). I didn't mean to rush, I was simply hoping to elicit a response=D. Smallman12q (talk) 21:54, 11 August 2009 (UTC)

Hello Smallman12q. I hope you don't mind if I butt in with a comment, well perhaps a question or suggestion, while we're waiting for comment on process. It applies equally to other applications for permissions on this page. Personally I find that Roux has hit the nail squarely on the head in the aforementioned other discussion.[4] The village pump can be slow sometimes, but I wonder if there's a lack of further comment because the initial response was so blindingly obvious. What I would like to see is a bigger case being made for why it would be an advantage to grant this right. Many of the things you mention don't require permissions, and I would expect some proven experience in them before applying for the right. Cobi for example (now a sysop) has bots with millions of edits and a high level of trust. The other non-sysop with the permission that I know of has specialist knowledge of a particular sockpuppeteer targeted by a filter, as evidenced by active support from a number of admins. I just haven't seen it from anyone else yet. There seems to me to be plenty of sysops around to edit the filters. You must accept there is some risk to granting this right to someone who hasn't been through the same scrutiny as a sysop. Where is the advantage to balance this risk? -- zzuuzz ^(talk) 23:19, 11 August 2009 (UTC)

I appreciate your remark: I always prefer discussion to silence=D. I do agree with Roux; abusefilter (and sysop tools) require a high level of trust by the general wikipedian community. An abusefilter, like sysop tools could be seriously misused if in the wrong hands (essentially blocking all edits). While documentation(which I have started) does not require permission, optimization and the creation of new filters does.

There isn't a set criteria; hence the thread at the village pump. As for advantage, that's where the wikipedia community comes in. As with an RFA, a consensus should be reached to determine whether granting such permissions is worth the risk(that's how I see it). The rationale behind that conensus should be held to some kind of standards...as yet to be determined.Smallman12q (talk) 00:01, 12 August 2009 (UTC)

It's been a few days...so...Smallman12q (talk) 14:41, 18 August 2009 (UTC)

Patiently awaits response...Smallman12q (talk) 14:56, 19 August 2009 (UTC)

This request is currently unsupported as far as I can tell. I'd suggest continuing to gain experience with regex and perhaps draft proposed filters (or suggest changes to existing filters) on this page for the next little while. I'm sure that after you have demonstrated your ability to write error-free filters people would be willing to support granting you this userright. (There is of course, the other path to gaining this userright you may wish to explore =) –xeno^talk 15:01, 19 August 2009 (UTC)

=(. I was looking forward to work on improving the edit filter.(While rfa may be an option, I don't really have a use for admin tools other than the edit filter).Smallman12q (talk) 00:48, 20 August 2009 (UTC)

I think you should also get an understanding of what warrants a filter: it has to be (in pseudocode): (frequent|malicious)&(difficult to prevent otherwise)&(represent consensus or other uncontroversial things, e.g. anti-vandalism). For instance, Wikipedia:Edit filter/Requested#New User Imposed Restrictions doesn't seem like a good choice for a filter because there is nothing preventing new users from prodding/deprodding articles. Moreover, even if we had such a filter, checking for "ipblock-exempt" and "sysop" in user_groups is unnecessary since they are all autoconfirmed (technically, they don't have to be, but in practice they are). -- King of ♥ ♦ ♣ ♠ 01:02, 20 August 2009 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

A functionality thought

It occurred to me (noticing a comment added to a filter which I didn't notice for two weeks) that it might be useful if filters could be watchlisted, so that (especially with the more obscure filters) it's easier to follow changes without having to regularly check the page. Has this been brought up before, or are there any major technical reasons why this wouldn't be possible to add? If not, I'll file a request... Shimgray | talk | 15:42, 19 August 2009 (UTC)

I don't think the two systems are compatible... –xeno^talk 17:22, 20 August 2009 (UTC)

Special:AbuseFilter/103 not working

Special:AbuseFilter/103 doesn't appear to be working properly. I tested it on new accounts twice[5][6] on Placodiscus caudatus (edit | talk | history | protect | delete | links | watch | logs | views). This filter hasn't picked up any false positives so far so should still be enabled.--Otterathome (talk) 18:46, 14 August 2009 (UTC)

Those don't match the filter criteria. –xeno^talk 18:49, 14 August 2009 (UTC)

Unless I'm misreading it, it is not testing for the "word" uncyclopedia, but for LINKS to uncyclopedia, so your edits would not have tripped it. -- Avi (talk) 18:51, 14 August 2009 (UTC)

Well it's suppose to filter it in any format.--Otterathome (talk) 19:29, 14 August 2009 (UTC)

This filter is more specific than you seem to think. –xeno^talk 19:33, 14 August 2009 (UTC)

It's private now so I can't see it, could you make the changes so it detects it in any format then? Thanks.--Otterathome (talk) 19:40, 14 August 2009 (UTC)

You want it to detect the word 'uncyclopedia' anywhere? Seems like the potential for false positives is too high. –xeno^talk 19:42, 14 August 2009 (UTC)

I set the filter public, no need for it to be private. Personally I disagree with this filter existing at all. Prodego ^talk 21:38, 14 August 2009 (UTC)

It seems you edit filter is not as effective anymore due to all the tweaking. If Oscar Wild and http://uncyclopedia both need to be detected at the same time then it will miss out a lot of spam as shown in the filter log. Prodego you've been against this filter from the start, it has filtered over 200 cases of vandalism without any false positives, so I don't understand why you are against it.--Otterathome (talk) 10:04, 16 August 2009 (UTC)

No, the current regex makes it Oscar Wilde or Uncyclopedia. This was the most recent successful catch. -- King of ♥ ♦ ♣ ♠ 15:53, 21 August 2009 (UTC)

Being that the above is an actual quotation by Wilde, wouldn't that qualify as a false positive? -- Avi (talk) 16:39, 21 August 2009 (UTC)

No, Template:Q displays . The editor who added that is probably from uncyc, where the Q template is a quote template. So I guess King is quite pleased with himself.--Otterathome (talk) 18:34, 21 August 2009 (UTC)

It still isn't sensetive enough[7], it should warn any user that inserts any new instance of 'uncyclopedia' in to an article. If you look at the history of vandalism the filter has prevented before you changed it you will see there are many examples of vandalism that the filter will no longer pickup, including[8][9][10][11][12][13][14] And that's only back to the 20th of July. I suggested before to have separate filters for this, and still feel it would be the best option to make everyones life much easier.--Otterathome (talk) 15:08, 22 August 2009 (UTC)

Filter Cleanup

Note that in a few hours I will be cleaning out various low hit / unused filters. Over 2% of edits are hitting the condition limit, and when that happens is when I tend to clear things out. I've listed some filters below that I may be disabling:

1. Filter #43 · hits
2. Filter #82 · hits
3. Filter #105 · hits
4. Filter #119 · hits
5. Filter #116 · hits
6. Filter #152 · hits
7. Filter #155 · hits
8. Filter #159 · hits
9. Filter #176 · hits
10. Filter #211 · hits
11. Filter #222 · hits
12. Filter #223 · hits
13. Filter #227 · hits

Prodego ^talk 17:50, 19 August 2009 (UTC)

As I side note I renamed various enabled test filters to 'User's test filter'. It makes it easier to spot them, and see who is responsible for it. If you create a test filter for yourself, please don't take over someone else's filter, they might want it. Either create a new one, or take over an old one you created. For small tests you can simply test on the active filter you want to change, after disabling all of the 'consequences' of a hit (warn off, disallow off, etc). Prodego ^talk 17:55, 19 August 2009 (UTC)

Thanks for keeping an eye on this. I've disabled 176 until I can spend some time completing it. —EncMstr (talk) 18:28, 19 August 2009 (UTC)

Hello. Filter 223 is important, even though it may not get used much. There are cross-wiki BLP defamation issues regarding articles titled under various forms of the name "Marco Lupis." I could go scrounging up the links to deleted pages if I need to, but I'd prefer not to make that big of a deal out of it. So yes, please do keep that filter. Plus it's only a couple weeks old. Keegan (talk) 20:34, 19 August 2009 (UTC)

Part of the problem may have been the 223 was written incorrectly, I fixed the syntax, we will see if it gets any hits now. Prodego ^talk 20:57, 19 August 2009 (UTC)

The filters above (excluding 223) have been disabled. Please let me know if you reenable any (but feel free to do so). Alternately, ask me to reenable them for you, if you can not. Prodego ^talk 03:04, 20 August 2009 (UTC)

And the cleanup was a success, condition limit hits went from ~3.3% to ~0.05%. Prodego ^talk 03:36, 20 August 2009 (UTC)

You only warned the users who made the filters yesterday, you could've waited a bit longer.--Otterathome (talk) 13:31, 21 August 2009 (UTC)

Could have, but didn't. Dragons flight and I have both done similar 'purges' before, without any 'warning'. Prodego ^talk 22:26, 21 August 2009 (UTC)

Would you agree that all filters all need to eventually be either set to warn or disallow, and the log only mode is for testing for false positives?--Otterathome (talk) 15:09, 22 August 2009 (UTC)

No. If someone is looking at the log, then the filter does not have to be set to warn, disallow, or tag. Prodego ^talk 00:12, 23 August 2009 (UTC)

Filters

I see we have managed to get up to 2% hits on the condition limit again. Looking at the changes since I cleaned out the filters, I see some totally inappropriate and unacceptable things, some terribly inefficient and ineffective things, and I simply can't do this all on my own. Everyone adds and adds and adds, and no one is removing except me, and I simply can not keep up with you all. Every filter needs to be reviewed. Please THINK before doing things, and do NOT use filters to block links. Prodego ^talk 21:31, 26 August 2009 (UTC)

Maybe if you and other admins constantly changing filters would actually try and discuss the changes instead of borderline edit-warring on Special:AbuseFilter/203 and have only said 1 thing each at Wikipedia_talk:Edit_filter#Special:AbuseFilter.2F103_not_working yet leave 'notes' on the filter page which can only be edited by admins. I find it very unbecoming that both admins have failed to discuss it though have carried on editing it over disagreements. Just look at the history. You ask to seek consensus for that particular filter, has it had any complaints or false positives? No. Do explain your reasoning? No.--Otterathome (talk) 19:21, 27 August 2009 (UTC)

I am merely trying to address problems pointed out by Prodego, refining it each time. -- King of ♥ ♦ ♣ ♠ 22:05, 27 August 2009 (UTC)

Special:AbuseFilter/7 is a very expensive filter. I tweaked it to run only 1/3 of the time. -- King of ♥ ♦ ♣ ♠ 22:08, 27 August 2009 (UTC)

Is Special:AbuseFilter/218 being used? It seems to be just taking 7+ ms without any actions. -- King of ♥ ♦ ♣ ♠ 00:27, 28 August 2009 (UTC)

Note the condition limit and the run times are two separate limitations. That said you can probably disable that one. Prodego ^talk 00:29, 28 August 2009 (UTC)

Special:AbuseFilter/231

I don't understand: Why is this catching URLs? I put in !(count("http://", added_lines) > count("http://", removed_lines)). -- King of ♥ ♦ ♣ ♠ 03:42, 28 August 2009 (UTC)

The only way to do that is to ignore any section that contains a URL. A better way would be to use rcount, but that is more intensive. I can help you with it later. Prodego ^talk 03:56, 28 August 2009 (UTC)

(ec x2) Which edit triggered it differently than what you want? What was the reason for comparing the number of added URL lines to the number of removed URL lines? I would have though you would just want to see if any HTTP: lines were added. As it stands now, if the users adds one HTTP lines and removes two then this "!>" will evaluate to 1. If the user adds one valid URL (which often contain these repeating strings) but deletes two old URLs the filter will still run.  7  04:00, 28 August 2009 (UTC)

But there are too many URLs floating around, especially in well-sourced articles (making false negatives too likely). Dang, I wish there was a function that could actually find regex matches rather than merely test for their existence or count them. -- King of ♥ ♦ ♣ ♠ 04:21, 28 August 2009 (UTC)

Priority of filters when multiple EF would be activated by a single edit

Apologies in advance if this has been covered. Searched the archive on "multiple" and didn't find it.

Has there been any consideration of (or does the community think there is any merit in) modifying the edit filters to run sequentially or with some sort of priority given? Based on this edit it would appear that once any edit filter is activated and the edit is tagged other edit filters do not run (understandably, probably for performance purposes). This case was tagged as "Shouting", but the EF for "Speedy Deletion Template removed" was not triggered and/or not tagged.

It would seem to me that the removal of a speedy deletion tag by an author should take a higher priority than a shouting tag. Assuming it's not plausible to continue to run all filters and tag with all possible violations I would like to suggest that it would make sense for us to assign a priority to each edit filter to decide which ones run first, to always ensure that the most severe tag (as agreed by abusefilter users) always was displayed.  7  05:02, 25 August 2009 (UTC)

They all run, and single edits can hit multiple filters. All consequences are applied (meaning you can get two warnings if you hit two filters set to warn). Prodego ^talk 13:09, 25 August 2009 (UTC)

Good - thanks - but any idea why this edit only had the shouting caught, not the Speedy tag removal?  7  13:35, 25 August 2009 (UTC)

Because filer 224, the copyvio tag removal filter, only looks for {{Copyviocore}} and {{csb-pageincludes}} being removed, and that edit removed {{db-copyvio}}. Prodego ^talk 13:39, 25 August 2009 (UTC)

Oh, and it didn't match the speedy removal filter, 29, because it does not apply to autoconfirmed users (which that user is). Apparently someone decided non-admins can remove speedy deletion tags if they object to them now. Prodego ^talk 13:43, 25 August 2009 (UTC)

Filter 29 was the one I was expecting to see I guess. Hmm... I agree that anyone can remove a speedy, but was hoping there was a way it would catch when the original author removed speedy from an article they created as in this case. Thanks anyway.  7  13:45, 25 August 2009 (UTC)

I think that autoconfirmed isn't a high enough level - maybe we should go for 50 edits or something? עוד מישהו Od Mishehu 11:54, 26 August 2009 (UTC)

Agreed.  7  22:54, 27 August 2009 (UTC)

 Done King of ♥ ♦ ♣ ♠ 06:04, 29 August 2009 (UTC)

EF Managers Permission

I've done some work with regex, and I also have some experience with programming languages like JavaScript (advanced beginner). I feel that I would be able to contribute positively to the Edit Filter, and I would discuss any major changes or questions I had on talk before enacting said changes. I am also open to any questions or "testing" to evaluate my candidacy/competence. Regards, MacMed^talk_stalk 18:36, 10 August 2009 (UTC)

Q: According to your userpage, you've been on Wikipedia for just shy of 3 months. EFM is a powerful permission that requires both technical knowledge and a refined understanding of Wikipedia's editing policies. Can you explain why you feel you meet both these requirements? –xeno^talk 18:41, 10 August 2009 (UTC)

A: I feel that I have the technical knowledge from dealing with some programming situations as well as working with Wikipedia regex in the past. Again, I am willing to answer/solve any examples or problems you may wish to present me with. As for the editing policies, I don't know if there is a way to prove that to you, but I shall try. Personally, I believe that a Wikipedian should be judged on the maturity and usefulness of their edits, rather than their wiki-age. I have made mistakes in the past, but once I was informed of the mistake I did not perform that same mistake again. Some of our current policies are, to me, self-explanatory and common sense (ie. WP:NPA, WP:AGF, WP:BITE, etc.). Others, like the notability policy and it's various subpolicies are more complex, but can be understood and taken to heart. I am not just dropping numerous links here, I can define and give you a description of each policy if you wish. Finally, as I said above, any questions I have, or actions I am unsure about, will be discussed either here or on the relevant notice/discussion board before the action is taken. Regards, MacMed^talk_stalk 18:50, 10 August 2009 (UTC)

Thanks. I must admit, in general I am apprehensive giving this right out to folks with such a short history. –xeno^talk 18:54, 10 August 2009 (UTC)

Q: Could you come up with a net-new filter that you might work on were this request successful? You may describe it and/or show the code. –xeno^talk 18:54, 10 August 2009 (UTC)

A: The first thing I would probably do is continue to help out User:Otterathome with the Uncyclopedia vandalism he seems to be experiencing. That coding is simple. Another idea I have is to help out with the seemingly common problem at WikiProject Biographies, which is the creation of biographic articles, which are then tagged with the {{WPBiography}} template, without the |listas= parameter. The code could get a little complicated, but I think I could handle it. A start could be:

(article_namespace = 1) &
(lcase(added_lines) rlike "{{(WPBiography|BIO|WPBIO)") &
!(lcase(added_lines) contain "|listas")

Not sure if there is already a filter for that, but it's an idea. The filter would log and warn, with a warning template with something like "You recently tagged an article with a WikiProject Biography tag, but did not include a |listas= parameter. When you can, please return to the article and include the parameter. Thank you." I would also keep an eye on requested filters, and any requests I receive personally, while helping with team collaboration when necessary. Regards, MacMed^talk_stalk 19:08, 10 August 2009 (UTC)

Would you care to comment at Wikipedia:Village_pump_(policy)#Criteria_for_Abusefilter.2Feditfilter_permissions for helping devolp criteria for granting nonadmins permissions?Smallman12q (talk) 21:42, 10 August 2009 (UTC)

It baffles me why anyone would consider failing to add a parameter to a maintenance template to be abuse. Gurch (talk) 23:31, 13 August 2009 (UTC)

That's why the name was changed to "edit filter" :). This would not disallow the edit, nor would the warning template be derogatory or accusative. It would just say "We noticed that you forgot to add this parameter. If you could, would you mind adding it? Thanks," It wouldn't disallow the edit or look negative on the editor. Just a simple reminder. Regards, MacMed^talk_stalk 00:02, 14 August 2009 (UTC)

Changing the name of something doesn't change its effect, nor the fact that it is being misused outside of its designed purpose. Gurch (talk) 21:58, 16 August 2009 (UTC)

Is this closed, or what? MacMed^talk_stalk 23:11, 30 August 2009 (UTC)

Request to remove Edit Filter access

A few weeks ago I requested edit filter access in order to help with identifying Scibaby socks. I will no longer take any part in identifying or reporting Scibaby socks, so there is no need for me to have edit filter access. Thanks to all the Edit Filter crew for your good work. Short Brigade Harvester Boris (talk) 06:09, 29 August 2009 (UTC)

I've removed the flag per your request. Let us know if you want it back, and thanks for your work. Shimgray | talk | 23:54, 30 August 2009 (UTC)

Warn action

It would be helpful for the filter log to specify if the user, having been warned about an edit, went ahead with the edit anyway. Evil saltine (talk) 02:08, 31 August 2009 (UTC)

If (s)he did, then there will be an other log entry for the edit. See, for example, here and here. עוד מישהו Od Mishehu 08:04, 31 August 2009 (UTC)

Special:AbuseFilter/9

Per Wikipedia:Edit filter/False positives#67.232.93.56, I added in an extra check to make sure that the personal attack is not in quotes. What we now end up with is a 9+ ms filter with three regex checks. It would be ideal to combine the quote check into the original added_lines check. I tried the following, to no avail:
lcase(added_line) rlike "\n[^\x22^\r^\n]\b(?<!it)(is|\'s|are|\'re|u\s*r\b)\s+(a|an|)\s*(fuck(ing|)|(u|ü)ber|ultra|very|)[ -]*((ass|arse)(hole|)\b|bastard|cock(sucker|s|)\b|cunt|d(1|i)ck|douche|fag|ghey|loser|kike|(mother|)fuck(er|face|tard)|nigg(er|a)|pa?edo(phile|)|prick|shit|slut|tosser|twat|wanker|whore|bitch|cabr(ó|o)n|culo|gay|idiot|moron|pendejo|penis\b|put(a|o))"
My idea was to search for a new line, and then everything leading up to the attack must not be line breaks or quote symbols (ASCII hex value = 22), but it doesn't work apparently. 1) Why doesn't it work? 2) Any suggestions? -- King of ♥ ♦ ♣ ♠ 23:54, 31 August 2009 (UTC)

Doesn't it need an asterisk after [^\x22^\r^\n]? Evil saltine (talk) 10:39, 3 September 2009 (UTC)

Syntax

What is the syntax for "contains" as opposed to removed or added? Rich Farmbrough, 07:31, 3 September 2009 (UTC).

Try old_text, new_text, old_wikitext, new_wikitext, old_html, or new_html. I was under the impression that one or more of these had been disabled due to bad performance, but you won't know until you try it. mw:Extension:AbuseFilter/RulesFormat. -- zzuuzz ^(talk) 10:21, 3 September 2009 (UTC)

End of article

Anything added to the end of the article that is not a stub template, interwiki or category is almost certainly needing attention. I would say about 1/3 each useful info, comments from people who don't "get it", and vandalism. An edit filter would be good if there was good chance of it being acted on. Rich Farmbrough, 08:35, 3 September 2009 (UTC).

Special:AbuseFilter/163

Is there a particular reason this isn't set to disallow? Protonk (talk) 20:01, 4 September 2009 (UTC)

In case someone creates a redirect first? Prodego ^talk 23:17, 4 September 2009 (UTC)

Is there a way to look at all edits triggering a specific filter (the tag created by filter 163 is shared by 2 other filters) without bringing up the large details interface? I want to do a check to see what percentage of the last 200 hits to the filter were not redlink redirect vandalism, but it will take forever if I do it by clicking through them on the log (and I'm hopeless w/ coding a bot to do it). Protonk (talk) 06:36, 5 September 2009 (UTC)

I'm surprised the following edit didn't trigger a filter

Shouldn't a your-mom edit have a filter triggered or something?192.12.88.7 (talk) 04:30, 5 September 2009 (UTC)

I turned it off. There would obviously be legitimate reasons for saying "your mom," so I am currently in the process of finding a way to catch bad instances of "your mom" while allowing the good instances. -- King of ♥ ♦ ♣ ♠ 04:40, 5 September 2009 (UTC)

Oh. And, by the way, here's another one (Filter 9 only protects against new or unregistered users or something). "Rohmaya Jackson" was a name made up for the occasion, so I doubt anyone was actually dissed by that. 192.12.88.7 (talk) 04:56, 5 September 2009 (UTC)

path name

although this has been renamed as Edit filter, the path that shows up when one uses special:pages is still http://en.wikipedia.org/wiki/Wikipedia:Abuse_filter and similarly http://en.wikipedia.org/wiki/Wikipedia:Abuse_filter/False_positives . Can this be changed also? DGG ( talk ) 18:27, 12 September 2009 (UTC)

It will always be Special:AbuseFilter. Prodego ^talk 21:37, 12 September 2009 (UTC)

DGG, those links will always remain as redirects to the current pages, just as the no longer user Image: namespace automatically redirects (As a pseudo namespace) to the new File: namespace. Its mainly for users convenience and nothing else :). Best, Mifter (talk) 15:12, 13 September 2009 (UTC)

Filter 149

Suggestion for Filter 149: exempt the case when the user name match the article title. Most of the hits is simply because of this. Dy yol (talk) 00:47, 14 September 2009 (UTC)

Seems like a good idea, provided some other filter will catch these cases. I donj't think there is one currently (there is a filter to catch creation of such pages, but not edits.) עוד מישהו Od Mishehu 08:23, 14 September 2009 (UTC)

There is a bot (COIBot) that catch this. Dy yol (talk) 16:10, 14 September 2009 (UTC)

I do not think this is necessary. Ruslik_Zero 19:06, 15 September 2009 (UTC)

Edit Filter 33 (talk page blanking) triggered upon creation (not blank) of the talk page.

Seems strange that this creation of the page would trigger the filter.  7  03:49, 17 September 2009 (UTC)

Would have thought "(old_size > 150)" would make it not worry about pages that didn't exist prior to the edit.  7  03:51, 17 September 2009 (UTC)

 Fixed: see the section below «Newarticletext in old_wikitext». — AlexSm 15:00, 17 September 2009 (UTC)

Bug: Newarticletext in old_wikitext on page creation

Apparently there is a new bug after recent MediaWiki update. When the page is created, old_wikitext instead of being blank contains whatever html MediaWiki:Newarticletext is producing in that namespace. Example: triggered. I would recommend disabling some filters right now until this is fixed. The same thing was also reported in the previous section. — AlexSm 04:27, 17 September 2009 (UTC)

Fix committed and deployed. — Werdna • talk 09:59, 17 September 2009 (UTC)

Thank you. — AlexSm 15:00, 17 September 2009 (UTC)

Blanking tag is buggy

There's something wrong with the Blanking Page detector. It's "filtering" newly created talk pages as "blanking". This is patently impossible. 76.66.196.139 (talk) 09:58, 17 September 2009 (UTC)

 Fixed: see the section above «Newarticletext in old_wikitext». — AlexSm 15:00, 17 September 2009 (UTC)

Request for a new filter

Can someone make a filter to detect edits where the word "India" or "partition" or "partition of India" or "partition of British India" is removed and replaced with "independence of Pakistan" or "independence" or "Pakistan"???? Some arbcom banned Pakistani users like to get rid of the word India at every opportunity. IT is quite a common trait. Thanks YellowMonkey (bananabucket) 06:12, 7 September 2009 (UTC)

I created Special:AbuseFilter/19. Ruslik_Zero 12:06, 7 September 2009 (UTC)

Ok, thanks, so how do I access the output to this, does it create a log of diffs, like AlexNewArtBot? YellowMonkey (bananabucket) 00:44, 8 September 2009 (UTC)

@Ruslik, please don't reposses old filters for new uses (unless it is one you created as a test filter). It messes up the log, and when we review new filters, we look at the newer numbers, so lower numbered ones might be missed. Low numbers + low hit counts tend to get disabled without anyone looking at the reactivation date. Prodego ^talk 01:30, 8 September 2009 (UTC)

Sorry, I thought it was better to reuse some filters disabled long ago. The current system is somewhat unsustainable—the number of filters continues to grow. Ruslik_Zero 08:21, 8 September 2009 (UTC)

AFAIK there is no upper limit on the number of filters, so it should be sustainable. There is a limit on how many can be active at once though. Prodego ^talk 17:50, 8 September 2009 (UTC)

How do I get it to write down a log of edits satfying this criteria on a page for reading? YellowMonkey (bananabucket) 05:09, 17 September 2009 (UTC)

The filter has caught nothing so far. Do you have any specific examples of edits that should be monitored? Ruslik_Zero 11:13, 19 September 2009 (UTC)

Special:AbuseFilter/246

I'm just wondering...what's this filter all about? The title says "Imma let you finish"... new meme and it's private. I hope we're allowed to know what is the filter about. 65.92.125.39 (talk) 02:37, 22 September 2009 (UTC)

Something to do with 2009 MTV Video Music Awards#Kanye West controversy and debated incidents. There is absolutely no reason for this abuse filter to be private. Gurch (talk) 12:55, 25 September 2009 (UTC)

Performance

Was the ms tracking disabled at Wikipedia:Edit_filter/Performance? It isn't showing anything for the last week, so I am guess it is disabled. I can understand why is might have been as tracking performance slows down performance and all... on the other hand it makes it difficult to concretely says "the filter is faster now that I made X change" (or conversely "oops, it is slower now"). --ThaddeusB (talk) 05:17, 27 September 2009 (UTC)

No, the bot is just broken. The information can be found on the filter description page. Which is fine for you lot, but us mere mortals can't view the description page for private filters, so the information is no longer available to us. Gurch (talk) 10:20, 27 September 2009 (UTC)

Filter 225 lying about its purpose

Filter 225 bears the terse description "Edit summary vandalism III". Since you don't feel like sharing the details with me, I can't tell what it actually does, but I can look at the log, where it is disallowing revisions with no summary or obviously benign summaries. It is thus obviously looking at the page text as well. Please change the description to describe what it actually does. -- Gurch (talk) 22:01, 27 September 2009 (UTC)

I renamed it. -- King of ♥ ♦ ♣ ♠ 22:28, 27 September 2009 (UTC)

No need for it to be private either. Prodego ^talk 01:22, 28 September 2009 (UTC)

Filter 249 assuming bad faith

Apparently, helpfully reverting vandalism is now a bad thing. Sometimes it amazes me that we still get contributions. Gurch (talk) 22:07, 27 September 2009 (UTC)

It doesn't warn the editor or prevent the edit, it just leaves a tag in the edit history. I don't think a bit of higher scrutiny on an edit like this is unwarranted.—Kww(talk) 22:15, 27 September 2009 (UTC)

Abuse Filter 245

Special:AbuseFilter/245 has gotten 7 hits in the past two weeks[15]. Perhaps we should disable it until we need it again? NW (Talk) 22:30, 28 September 2009 (UTC)

Filter 131

Can someone please reenable Filter 131? Prodego disabled it without discussion, and 131 discussion with him on his talk page has not been fruitful. This was enabled by consensus of regular editors at the two articles it protects, Mohammed and Temple garment. These articles contain images that are considered blasphemous by Islam and the LDS, respectively, and are subject to people removing them without any regard to Wikipedia consensus. What Prodego sees as "only 76 hits in 150 days" can also be seen as "an edit war every other day".—Kww(talk) 01:16, 30 September 2009 (UTC)

No one is questioning that the edits it blocked were bad edits, but that isn't the point. That filter was adding 3ms to every edit across the entire encyclopedia. Given those stats the filter was adding nearly about 19 minutes of processing time for every catch. Given that, Prodego clearly made the correct call. --ThaddeusB (talk) 02:10, 30 September 2009 (UTC)

And FYI, there were in reality a lot less than 76 hits because a large chunk were the same user ignoring the warning and then being stopped when they did. The unique hits were less than 50 and about half the hits were on Rorschach test. Additionally, it had actually been more than a month since there was a hit on a religious article. --ThaddeusB (talk) 02:21, 30 September 2009 (UTC)

This filter was very usefull—Rorschach test is semi-protected now. Ruslik_Zero 11:53, 3 October 2009 (UTC)

Special:AbuseFilter/231

I think we should re-enable this filter, which regularly gets hundreds of hits a day. Yes, many of its hits are indeed caught by 135, but the repeating characters and unspaced characters are two distinct indicators of vandalism, and should be tagged as such. -- King of ♥ ♦ ♣ ♠ 02:52, 6 October 2009 (UTC)

Trying to copy to another wiki

I tried to copy the "New user blanking articles" filter to another wiki and got the following error: Unrecognised variable contains_any at character 141 Does anyone know what I need to do to fix it? Hintswen  Talk | Contribs  23:59, 6 October 2009 (UTC)

Abuse log -> Filter log

I noticed that some of the user information template, like {{IPvandal}} continue to use the term abuse log instead of filter log. Are there any objections to updating the terminology to "filter log" to be consistent across en.wiki ? I think this should be non-controversial, but am checking to make sure. Abecedare (talk) 04:23, 10 October 2009 (UTC)

No objections. Ruslik_Zero 08:05, 10 October 2009 (UTC)

Done. Abecedare (talk) 16:03, 10 October 2009 (UTC)

Warnings not indicating the filters using it

There are a few of them, I'll place them here for reference:

• MediaWiki:Abusefilter-warning-advertising
• MediaWiki:Abusefilter-warning-badmove
• MediaWiki:Abusefilter-warning-copyvio
• MediaWiki:Abusefilter-warning-helpme (found: Filter 104)
• MediaWiki:Abusefilter-warning-largeaddition-notice (incorrect filter)
• MediaWiki:Abusefilter-warning-priorcase

Note also that I have modified Template:Edit filter warning so that it can handle up to 5 filters, with filter2=, etc. The complete list of filter warnings is at Special:Prefixindex/Mediawiki:Abusefilter-warning. Cenarium (talk) 15:21, 10 October 2009 (UTC)

Filter 104 has a non-standard message. Note that this filter is marked as deleted. עוד מישהו Od Mishehu 11:12, 11 October 2009 (UTC)

I have deleted several warnings for deleted filters, no opposition to keep that way ? Cenarium (talk) 14:24, 11 October 2009 (UTC)

Filer 65 (white space) has many false positives

Someone pointed out on the wikimedia-tech IRC channel that filter 65 (the whitespace filter) has a lot of false positives. For example, just aligning parameters of a template can trip it. The filter right now is set to warn and tag. What if we change it to just tag, given that the false positive rate is so high? — Carl (CBM · talk) 18:55, 11 October 2009 (UTC)

I agree. I wanted to do this myself. Ruslik_Zero 19:00, 11 October 2009 (UTC)

The filter has some false positives certainly, but on the other hand if it doesn't warn there isn't much point. The idea is to warn users that they are using formatting (manual whitespace) that is likely to produce results that look funky, but aren't really wrong per say. I don't see a point in tagging only since the user might fix their error (or at least try to) themselves, and even if they don't I don't think people are actively seeking these formatting errors to fix.

Thus the filter should either to improved to an acceptable level of false positives, or turned off entirely. Looking at the actual filter a string of spaces followed by an equal sign shouldn't trigger it, so it shouldn't trip on just aligning template parameter. --ThaddeusB (talk) 21:35, 11 October 2009 (UTC)

This is what I meant: [16] and [17]. A different sort of alignment issue is shown by [18] [19] and [20]. A third issue is [21] and [22].

Personally, I think that there is little reason to merely warn users about this. If they insert whitespace that breaks things, they or someone else will fix it soon enough. If they add whitespace that doesn't break anything, who cares? If we were using this to stop edits, that would be one thing, but if we are going to let the edit through anyway, I don't see the point of the warning in this case. The log shows that most editors repeat the edit, which I think means they go ahead and save it anyway. So our warning is not helping them fix the formatting problems. — Carl (CBM · talk) 22:13, 11 October 2009 (UTC)

Prodego has disabled the warning for now... Personally, I am leaning towards disabling the filter entirely as I don't think it is of much use. However, I will take a closer look at the log and see what % of the edits are completely harmless & re-evaluate when I have more time. --ThaddeusB (talk) 02:15, 12 October 2009 (UTC)

Just tagging them seems at least hypothetically useful in a watchlist, to see that the edit might require reformatting even if the content is sound. — Carl (CBM · talk) 02:23, 12 October 2009 (UTC)

CAPTCHA's and the edit filter

After reading that thread on WP:VPT about the addition of external links by anonymous users, I had a thought. It seems like that being able to trigger CAPTCHA checks for anonymous users adding external links is good, but we could use some finer control over it. Is it possible, to use the edit filter system, to trigger CAPTCHA's for specific user groups when triggering a specific filter? ViperSnake151  Talk  12:05, 15 October 2009 (UTC)

Filter 50

Special:AbuseFilter/50

This edit diff is a false positive. The editor is spelling out a DNA sequence. I'm not sure if these are common enough to be a problem, but if you wished to avoid these you could exclude edits that only had A, T, C and G as the caps letters. Tim Vickers (talk) 16:42, 15 October 2009 (UTC)

This is one of those filters that's supposed to check "the same thing" either in removed_lines or in old_wikitext but it doesn't. Plus, it would probably be nice if the filter would check some other common conditions to prevent useless triggering, e.g. new_size < 50 when #3 New user blanking articles fiires anyway. — AlexSm 17:26, 15 October 2009 (UTC)

Reqest for pemission (Hintswen)

Not sure if this is the right place buy anyway... I've started my own wiki and now I've started to get vandalism. I'm looking at using the AbuseFilter extension for my wiki and so I want permission to see the hidden filters here so I can copy them over to my wiki. Hintswen  Talk | Contribs  01:35, 23 September 2009 (UTC)

To me this does not seem like a valid reason for this permission. Many hidden filters are wikipedia specific, and the reason they are hidden is to stop just anyone from inspecting or copying them, so as to stop the vandals that they are aimed at from seeing them. Martin451 (talk) 01:43, 23 September 2009 (UTC)

Could you give us a link to the wiki that you want the extension for? MacMed^talk_stalk 01:52, 23 September 2009 (UTC)

HabboWiki.org is the wiki I am looking at installing it on. It's only new (made it at the start of the month) but it's aimed at an audience more likely to vandalize it.Hintswen  Talk | Contribs  02:03, 23 September 2009 (UTC)

This isn't a vote, so no Oppose or Support. Consensus in its truest form this is, because we don't have a million people here. Generally we wouldn't give EFeditor just so people can see private filters, however, this was done once before (for Charitwo, who also wanted to copy them over to his own wiki). The key thing here is that the PUBLIC filters tend to be the complicated ones, the private ones are the simple ones, the reason they are private is because they would be easy to get around. Prodego ^talk 01:58, 23 September 2009 (UTC)

I agree with Prodego here. I checked the recent changes for the wiki, and there is some vandalism, but most of it is basic profanity. AFAIK, the profanity filter is public. Again, as Prodego said above, most of the private filters are private because they are very simplistic and targeted at very exact kinds of vandalism (guessing here). It's really up to the admins here, but I personally don't really see the harm in letting Hintswen see the private filters. Maybe even agree to have the perm taken away after importing whatever filters he/she needs? Regards, MacMed^talk_stalk 02:21, 23 September 2009 (UTC)

Well I don't mind skipping the private filters. I'm guessing the public filters will catch most (if not all) of the vandalism for now and anything else I'll just have to make my own filters for. As I said, it's only a new wiki so there wont be much vandalism for now anyway. Just thought I'd ask for the permission so I could get the whole lot (for future vandalism). Hintswen  Talk | Contribs  02:30, 23 September 2009 (UTC)

If it's a specific one you want, someone can always email it to you. — Jake Wartenberg 16:19, 3 October 2009 (UTC)

•  Not done You don't need the edit filter manager to view public filters. –xeno^talk 17:49, 16 October 2009 (UTC)

Permissions for non-admins; discussion

The procedure for granting the edit filter permission remains unclear.

Wikipedia:Edit filter states;

• Presently, requests for assignment of the "Edit Filter managers" group to non-admins should be made at Wikipedia talk:Edit filter, where a discussion will be held for up to a week prior to a decision being made.^[1]

1. Process informally agreed upon when Cobi (talk · contribs), who was not yet an admin at the time, was granted the userright in this thread: Wikipedia talk:Edit filter/Archive 3#Abuse Filter editors group membership request.

I made such a request on this page, above, Wikipedia talk:Edit filter#Reqest for pemission.

I personally do not mind whether the permission is or is not available to non-admins - and I am no expert in the field, so I am not proposing a solution/policy. I merely think that it needs clarifying, so I welcome opinions from the community, to form a coherent policy or guideline on the matter.  Chzz  ►  23:31, 7 October 2009 (UTC)

I find it interesting that this, perhaps useful discussion, continues to be ignored. –Katerenka (talk • contribs) 22:14, 11 October 2009 (UTC)

Now Wikipedia:Administrators' noticeboard/Archive239#Edit filter permission  Chzz  ►  17:14, 16 October 2009 (UTC)

Struck, above, as - after raising the question on AN, I have been granted the permission; see earlier resolved request section. I still think that some discussion/clarification on this would be beneficial.  Chzz  ►  05:28, 17 October 2009 (UTC)

!autoconfirmed

Does placing a non-autoconfirmed user in the "confirmed" usergroup allow them to bypass edit filters that target non-autoconfirmed users? Someguy1221 (talk) 06:34, 19 October 2009 (UTC)

No, it doesn't. עוד מישהו Od Mishehu 09:46, 19 October 2009 (UTC)

Edits sneaking through. Filter system maxed out?

I see edits - from 90.200.240.27 (talk · contribs · WHOIS), e.g. - sneaking through filters - 258 in this case. What gives? Does that mean the filter system has been maxed out? Wknight94 ^talk 13:36, 20 October 2009 (UTC)

Yes, and if you want to help fix that go through any filter NawlinWiki has edited and remove old terms. There are huge numbers of old vandal phrases that he never removed. Prodego ^talk 02:16, 21 October 2009 (UTC)

I've pruned several filters. I haven't removed everything -- there are some catchphrases that some banned users just can't seem to get over (see the last couple of pages of the filter 58 log). NawlinWiki (talk) 02:41, 21 October 2009 (UTC)

Lets give the condition limit percentage a minute to stabilize, and see where we stand. NW, if you could make sure to reevaluate all the old terms before adding new ones, that would go a long way to reducing this sort of 'condition rot'. (growth of checks of limited usefulness) I can watch over new filters, and old ones that might no longer be worth it, but I can't keep track of specific terms in those big list filters. Prodego ^talk 02:54, 21 October 2009 (UTC)

One idea might be a write a program to analyze the filer logs to see which terms are actually catching. --ThaddeusB (talk) 03:36, 21 October 2009 (UTC)

Filter 58 and Mr.Z-Bot

For some time now, Mr.Z-bot (talk · contribs) has been monitoring the filters that may be accurately described as abuse filters and reporting the hits to AIV. For those who aren't familiar with it, this bot only reports after a number of hits to a filter that catches simple vandalism, but reports immediately after hitting a filter designed to catch certain repeat or copycat vandals. Filter 58 is an example of the latter. Well, I've been seeing quite a number of false positives at AIV caused by these two. While most of those false positives were indeed vandlism, they were not the sort of vandalism that would warrant an immediate block. I was wondering if it would be feasible to split filter 58 between phrases used by committed vandals, and certain more general phrases, or if it would make more sense to simply ask for the bot to distinguish such hits to the filter. I've asked the bot's owner to comment here. Someguy1221 (talk) 05:55, 24 October 2009 (UTC)

Splitting the filter would be far easier. Mr.Z-man 06:38, 24 October 2009 (UTC)

Alright. I've split off from 58 the three phrases that would only warrant a warning into Filter 260, and updated the bot's filter list accordingly. Someguy1221 (talk) 00:47, 25 October 2009 (UTC)

Which filter...

Just wondering, which filter detects Grawp-related page-move vandalism? (such as moving pages to H.A.G.G.A.R. variations?) Thanks. —Preceding unsigned comment added by 65.92.126.114 (talk) 13:52, 11 October 2009 (UTC)

Not much point in having private filters if we answer questions about them in public. EdJohnston (talk) 17:34, 17 October 2009 (UTC)

Filter 72 found at least some of them, although they might hit other filters too. It's not too hard to reverse engineer private filters if you check the logs and are willing to experiment a little. 76.211.11.31 (talk) 19:36, 27 October 2009 (UTC)

False positives in filter 58?

Can someone tell me what is going on here? On the morning of October 25, Patkenel (talk · contribs) was making some edits to 2010 Olympic torch relay route, and it seems that a sizable number of them were disallowed by the filter, claiming "attacks on specific users", but I cannot see anything wrong with any of the edits. Is this a bug? If so, I think it's a pretty major bug that should be fixed as soon as possible. I wish I could see the filter itself so I could make a guess at what the actual problem might be, but I'm not an administrator. -- Soap ^Talk/_{Contributions} 19:54, 25 October 2009 (UTC)

Link to logs: http://en.wikipedia.org/w/index.php?title=Special:AbuseLog&limit=120&wpSearchFilter=58

Zzuuzz fixed that 10 hours ago, thanks. Prodego ^talk 20:00, 25 October 2009 (UTC)

Okay, thanks. I had assumed it might still be a problem because it was later than the last discussion on this page. -- Soap ^Talk/_{Contributions} 20:07, 25 October 2009 (UTC)

Yep, you couldn't have seen the change, please do leave a note if there are any other problems. Thanks Soap! Prodego ^talk 22:32, 25 October 2009 (UTC)

Request for permission

Resolved

 – EFM permission granted to User:Chzz to be used to view only unless an explicit request to be allowed to edit as well is granted. Chzz is reminded not to discuss details of private filters except with other edit filter managers in a non-public venue. –xeno^talk 17:27, 16 October 2009 (UTC)

I'd like to be able to view the current edit filters; I don't intend to modify anything, but realize that the permission would enable me to do so; per the note "requests for assignment of the "Edit Filter managers" group to non-admins should be made at Wikipedia talk:Edit filter" I post this here.  Chzz  ►  02:43, 10 September 2009 (UTC)

If you don't intend to modify anything, I don't see the point. Prodego ^talk 04:01, 10 September 2009 (UTC)

(a) I trust Chzz with "view only".

(b) I don't think "view only" exists. I only see "edit filter manager", which is read/write.

(c) Prodego, the point is to see filters and logs that are normally hidden or locked. It was certainly a reason I wanted to get the admin bit. tedder (talk) 06:10, 10 September 2009 (UTC)

Why would you need to see the hidden filters if you aren't going to change any of them? All the hit logs are public, the only logs that are hidden for hidden filters is the history of the filter, not the hit log. What does Chzz intend to use the ability to view hidden filters to do, if not edit them? Prodego ^talk 12:14, 10 September 2009 (UTC)

Curiosity. In reality, every good editor who won't use the information to bypass the filters should be able to view it automatically. As we can't do this, I see no problem with giving it to a trustworthy editor who specifically asks for it. Ale_Jrb^talk 13:33, 10 September 2009 (UTC)

Right. It answers the "why did edit X get flagged, but edit Y didn't get flagged?". That's key information. Even if someone doesn't have the skills and/or the bit to change a rule themselves, they can help by researching the information, showing test cases, etc. tedder (talk) 16:42, 10 September 2009 (UTC)

That information is useless if you don't use it for anything, and I am not particularly interested in dozens of people requesting a user right they have no intention to use, simply so they can have it. "Researching the information, showing test cases, etc" can be done without the abusefilter manager right, and only for the 'private' filters can the "why was x caught and not y" not be answered. As to your comment Ale jrb, if you aren't going to do anything with a user right there is no point in you having it. 'What will you do as a sysop' has been question #1 on RfAs since the questions section was created. Prodego ^talk 17:56, 10 September 2009 (UTC)

Viewing the details of the filter is doing something with the right. And first, there are not dozens of requests - there's one. Second, you don't have to be particularly interested - any admin can grant it. There's no load on the wiki, and if the user is trustworthy and isn't going to change the filter, there's no real reason that they shouldn't have it - 'I don't like it' doesn't count. And finally, this isn't an RfA - arguably, however, having +sysop just so you could view deleted articles would be technically sufficient, even if it wouldn't get through RfA. Hell, someone with 50 edits passed just to edit a blacklist. Ale_Jrb^talk 20:06, 10 September 2009 (UTC)

That rationale is no better than wanting adminship so you can look at deleted pages, which as you say, would never fly. Any admin can, but may not grant the abuse filter manager right, that is why the instructions are to discuss here. There are only 2 non-admins with the abuse filter manager right. It is annoying enough with admins giving themselves the right and then doing nothing. Prodego ^talk 20:22, 10 September 2009 (UTC)

I am aware that admins shouldn't just go round granting the right. It's simply that your comments make it sound like you are personally irritated by the request because your workload will increase or something, which is completely irrelevant: if I've completely misread that, sorry! And that RfA would be most likely to fail on trust issues - I would suggest that a sufficiently trusted user could get through RfA saying they would only look at deleted pages; it would just never happen because such users are rare and generally want adminship to perform the tasks anyway. And many people also feel that we make too big-a-deal of +sysop. Anyhow, the current number of people with the right or what would happen in an RfA is irrelevant - this is a discussion of Chzz, here. And I see no real reason why not. And finally, if someone has the right but does nothing with it (take me, for example), in what way does that affect you? Cheers, Ale_Jrb^talk 20:33, 10 September 2009 (UTC)

I gave myself the permission to examine what was happening, not with any particular intent to modify things or test new ideas, though I may eventually do it also--I do know basic regex. It's a reasonable thing to want to do. For that matter, one of the main reasons i asked to become an admin was to check deleted pages to see if they could be rescued, and I said so at the time, and it was accepted as a perfectly good reason--and so I have been doing all along--though I've of course gotten into some other things also. This is a much less powerful privilege than sysop, and we shouln't be over-restrictive about it. DGG ( talk ) 18:27, 12 September 2009 (UTC)

• Q. Is there a certain private filter you're planning to look at, perhaps help with suggestions on how to improve it, etc? –xeno^talk 20:36, 10 September 2009 (UTC)

(EC) Blimey, I didn't intend to cause a big deal here! I'm really not bothered; don't worry about it. The reason that I asked was, I was looking at some rather persistent ongoing IP vandalism, from wide-ranging IPs over several articles. Some IPs have been blocked, and a couple of pages were protected. I had a chat with some others about it, and one thing I wondered was, if the edit-filter might be one way of tracking such a thing. I know that I could ask someone about it, but thought that if I could look at the existing ones, it might put me in a better position to suggest an appropriate filter or something. Note, this is not the "rationale" for wanting it, I'm just explaining how I came here, and it might serve as an example of why someone might want to look at these things. I expect I could spend hours thinking about all kinds of reasons for wanting to see it, but really - like RfA - it's simply not worth the trouble.  Chzz  ►  22:05, 10 September 2009 (UTC)

• Q. What's with the block log for goatse? -- zzuuzz ^(talk) 21:48, 10 September 2009 (UTC)

I was not going to answer this question, because I don't see the relevance, but I don't want to be accused of hiding anything - so; forgive me if the exact details are hazy, I can check them if necessary;

Some year-and-four-months ago, shortly after becoming a 'serious' contributor, I was full of the spirit of WP:BOLD, and embraced the freedom of Wikipedia. I thought that adding an image would make the article on 'goatse' much more informative, and I felt that the discussions on the topic did not show any policy-based reasoning that prohibited just going ahead and adding one. I uploaded the image, and it was almost instantly deleted, so with righteous indignation, I tried again, and a third time - and was blocked for just over 1 hour.

I certainly wouldn't do anything like that now that I understand things better; I would, instead, work towards consensus. I've made quite a few contributions since then, and not had any significant drama at all.  Chzz  ►  00:56, 12 September 2009 (UTC)

• Q. If you see an apparent typo in a hidden filter would you a) do nothing b) fix it c) ask someone on this page to fix it d) ask a friendly admin on their talk page to fix it e) ask someone on IRC to fix it? f) something else? -- zzuuzz ^(talk) 21:48, 10 September 2009 (UTC)

I certainly wouldn't be so naive as to post any portion of a hidden filter on-wiki, or indeed on any other public channel.

Nor would I edit it (as I stated in my initial request).

I would not approach any old 'friendly admin', as I am aware that not all admins have access to this area, and although they could grant it to themselves, many would not be familiar with the nature of the filters.

Most likely, some user that I knew to be involved with edit-filters would be active on IRC (identified with their password, etc) and I would ask them there about it, via a "PM" (personal message). Failing that, well, it would depend on the significance, urgency, and nature of the typo. It might be appropriate to put a note on this page, e.g. "There is a typo in edit filter number xx". Most likely is I'd wait until someone I knew to be appropriate was online; otherwise, I could email.

I am fully aware of the necessity for both discretion in discussions of filter contents, and in the need for caution in any change to them.  Chzz  ►  00:56, 12 September 2009 (UTC)

o-/ did I fail?  Chzz  ►  01:35, 23 September 2009 (UTC)

Yes, because you're trying to be useful, which seems to be frowned upon here in favour of DHS-style "if you're not one of us you're a terrorist vandal" paranoia. Gurch (talk) 15:41, 27 September 2009 (UTC)

• I've granted this request. The only objection was about the concept of granting an editing right for view-only purposes. Since the alternative below doesn't exist (and we're actually still waiting for it to be rolled into admin package, so I wouldn't hold my breath for a separate userright) and Chzz is a trusted user, the permission should allow him to assist in developing better filters. And if not, no harm done. –xeno^talk 17:36, 16 October 2009 (UTC)

An alternative

There is a right called abusefilter-view-private, that allows people to view private filters, but not modify them. This right is not assigned to any group now, but it is possible for it to be assigned to a new group, or it can be assigned to an existing group, like the rollback group. It's should also be assigned to administrators too. This might what you been looking for. Techman224^Talk 01:27, 29 September 2009 (UTC)

It's not available yet (No. 20721). -Jeremy ^{(v^_^v Stop... at a WHAMMY!!)} 11:01, 5 November 2009 (UTC)

Request for permission?

Hello. I would like to be granted the abusefilter right, if possible. I have a good understanding of Regex, and filter syntax, and I have my own Wiki on my PC I experiment with. Filter 249 was created at my suggestion, which routinely catches people like this [23]. I would like to help optimize existing filters and see if there are any other ideas I get as I do whatever it is I do normally. I have a clean block log and rollback enabled. Triplestop x3 02:14, 6 November 2009 (UTC)

False positives in Filter 146

Based on a few reports of false positives on the dedicated page, as well as a significant number of false positives I've removed from AIV, Filter 146 seems to be generating a substantial number of false positives. To avoid revealing anything about the filter, I'll just say that the problematic condition starts at around character 219. I'm not familiar with this troll, so I came here to ask whether this condition is really that necessary, before I remove it. Someguy1221 (talk) 02:27, 7 November 2009 (UTC)

Tweaked. Feel free to contact me offline. Wknight94 ^talk 02:32, 7 November 2009 (UTC)

Deleted edits survive in the Abusefilter logs

A recent edit by a vandal claims to show the real life name and location of another Wikipedia editor who has been a target of the vandal's attacks for some months now. This makes me wonder if some people who would want their edits to survive, particularly if they contain WP:OUTING violations, might trigger abusefilters on purpose just to take advantage of this loophole. Are there any plans to close off viewing of abusefilter logs, either selectively or entirely, to prevent the deleted edits from surviving? -- Soap ^Talk/_{Contributions} 03:53, 15 November 2009 (UTC)

This is T20043. Cenarium (talk) 04:08, 15 November 2009 (UTC)

The Necessity for Private Filters

Is it true that private filters hide personal informations about IP addresses of specific vandals, and hide sensitive information, such as the types of vandalism that vandals do? If that's the case, I'm wondering, does Filter 12 and 34 need to be private? Especially #34 because this filter's title is very straightforward about what it does, and #12 has a similar situation. —Preceding unsigned comment added by 65.92.126.254 (talk) 02:15, 23 October 2009 (UTC)

34 needs to be private. 12 less so, but a case could be made for it. But for the sake of beans (or something like that), I'm not going to say why they need to be private. Someguy1221 (talk) 02:20, 23 October 2009 (UTC)

I see...so that the vandals might be tempted to try doing the actions prohibited by filters 12 and 34 if it's made public? —Preceding unsigned comment added by 65.92.126.254 (talk) 02:28, 23 October 2009 (UTC)

More like, if you told the vandals how the filter worked, then it wouldn't work. Someguy1221 (talk) 02:33, 23 October 2009 (UTC)

Yep, I understand now, that if they were public vandals would find ways to bypass the filter by not reaching the criteria of the filter by writing words not covered by the filter. Also, should 123 be private as well? It displays the number of edits the user need to make to bypass the filter. 65.92.126.254 (talk) 02:37, 23 October 2009 (UTC)

123 was originally private, but then made public. I assume the change was based on the fact that there are no vandals that try this is any extreme excess, as opposed to pasting obscentities on pages. In other words, there's no point making a filter private if no vandals would try to find a way around it even if it were public. Just don't stuff any beans up your nose. Someguy1221 (talk) 02:52, 23 October 2009 (UTC)

Typically, the private ones are simple lists of things you can't do, which, if you knew what were, you would easily be able to get around. I agree with Someguy in that 34 needs to be (specific cutoffs) and that 12 less so, but still enough that I won't change it. Prodego ^talk 03:51, 23 October 2009 (UTC)

Sigh. Security through obscurity never really works out in the long run... --Tothwolf (talk) 08:57, 23 October 2009 (UTC)

I disagree. Prodego ^talk 11:28, 23 October 2009 (UTC)

MediaWiki:Titleblacklist (which is open to public viewing) seems to be working just fine. Anyone with a basic understanding of regular expressions can figure out how most of the "private" edit filters work. --Tothwolf (talk) 11:54, 23 October 2009 (UTC)

Again, I disagree. Prodego ^talk 04:13, 25 October 2009 (UTC)

It may not work well in the long run, but its currently our only practical option, and it works a whole lot better than than no security at all. Given that we're a website that anyone can edit, security by design isn't possible, as the "security hole" used by vandals is the "feature" that allows constructive edits. The title blacklist was almost completely ineffective at stopping dedicated vandals. Mr.Z-man 21:20, 25 October 2009 (UTC)

While I'd very much love to prove to Prodego just how easy is is to determine how the majority of these filters work and write up very close approximations of the rulesets that they use, I know of no non-disruptive methods to probe the edit filters and doing so would certainly be a violation of WP:POINT so I can't in good conscience do so. A "dedicated" vandal won't have any such morals though and without a doubt will be able to figure this stuff out. In the long run, hiding things from view is simply not going to work and may very well end up enticing vandals to be even more disruptive as they probe at and attempt to work around edit filters. I've also noticed a trend with these filters in that more and more of them are being set private or flagged as private when they are created. At the very least, hiding this stuff from the community should be strongly discouraged as it tends to imply that the community simply doesn't need to know or is somehow better off in the dark. --Tothwolf (talk) 20:07, 27 October 2009 (UTC)

Tothwolf, I'm not sure that would accomplish anything, but I'm sure that we could set up a filter on a test wiki and see if you can figure it out. It isn't quite so trivial. Prodego ^talk 03:05, 5 November 2009 (UTC)

I'd just like to note that any IP addresses in the range 65.92.*.* posting on this page are likely to be User:ScienceGolfFanatic and any admin can search the deleted edits as well as the "disallowed" abuse filter hits from this IP range for evidence. it would seem that he's trying to work out the information in the hidden filters by making masses of edits specifically intended to trigger them. I would have known this right away, but this page wasnt on my watchlist until a few days ago. -- Soap ^Talk/_{Contributions} 18:48, 15 November 2009 (UTC)

How are Abuse filters imported?

How can abuse filters be imported to another wiki? I'm an administrator in a non-Wikimedia-related wiki which has had a great deal of pagemove vandalism for the past couple of months. —Preceding unsigned comment added by 65.92.126.114 (talk) 20:26, 15 November 2009 (UTC)

Im pretty sure that this is an indef-blocked user who just wants to know more about the abuse filter so that he can figure out how to evade them. -- Soap ^Talk/_{Contributions} 00:14, 16 November 2009 (UTC)

I honestly fail to see the logic of how figuring out a way to export this program to another wiki would be a way to evade blocks. Exporting the filter and figuring out how to evade blocks are two totally different things. And I'm just someone looking for help in a legitemate way. 65.92.126.114 (talk) 01:29, 16 November 2009 (UTC)

The above user appears to be a sock of "ScienceGolfFanatic"; cf. [24]. I believe he is trying to figure out how to trigger as many filters as once for disruption. Triplestop x3 01:39, 16 November 2009 (UTC)

My question is how to export this program to another wiki. That's not trying to trigger multiple filters, because if I import the program to another wiki the intention is to benefit another wiki and the intention would not be to harm this wiki. 65.92.126.114 (talk) 01:42, 16 November 2009 (UTC)

Install the mw:Abusefilter program on your Wiki and write filters to stop whatever problem you have. Triplestop x3 01:50, 16 November 2009 (UTC)

The page says that it requires a local settings folder for that wiki in order for it to be implemented. I'm curious if it means that only the owner of that wiki would be capable of installing it. 65.92.126.114 (talk) 02:07, 16 November 2009 (UTC)

Yes, you need access to the server. Or you could just have a bot ban anyone making certain bad edits. Triplestop x3 02:15, 16 November 2009 (UTC)

Okay, thanks. 65.92.126.114 (talk) 02:26, 16 November 2009 (UTC)

Repeating characters

I've found "repeating characters" (#135) to be one of the most useful filters. I've reverted lots of vandalism and test edits using it. I hope that whatever issues led to its being disabled will be resolved. MANdARAX • XAЯAbИAM 22:50, 15 November 2009 (UTC)

Thanks, Wknight94, for enabling it again. I just reverted a batch of vandalism which I found using this filter. MANdARAX • XAЯAbИAM 07:17, 18 November 2009 (UTC)

Please consider adding "party pooper" to the list of exceptions on Filter 46

Currently it only lists

contains_any(lcase(added_lines), "poop deck", "poopó", "poop alley studios", "poo power", "come poop with me", "http://www.stationstops.com/2008/03/25/nyc-public-restrooms-the-straight-poop-on-public-restrooms-in-new-york-city-and-beyond/", "http://poopthebook.com/blog/2007/08/14/paradox-public-bathrooms/"

I think "party pooper" should be added as it has recently disallowed a legitimate edit that needed to insert that phrase. See this log. -- Soap ^Talk/_{Contributions} 18:48, 17 November 2009 (UTC)

 Done, since this has come up twice now. Triplestop x3 03:14, 27 November 2009 (UTC)

Atlantic Records

I think it would be a good idea to create a filter that prohibits replacing the article on Atlantic Records with "YOU SUCK!". If you look in the page's history, you will see that this is done very often, because Weird Al Yankovic does this in his music video "White and Nerdy". Draftydoor (talk) 15:38, 25 November 2009 (UTC)

You say "very often", but I see an average of less than once per month. The abuse filter runs on every single edit so that rate of abuse does not warrant a filter. Now if the filter could be fine-tuned to situations like this, that would be nice... Wknight94 ^talk 15:51, 25 November 2009 (UTC)

Yeah. Checking out the history again, I see that even though a large proportions of edits to this article are "YOU SUCK" vandalism, they do not happen all that often, so a filter may not be warranted. On the other hand, the article is currently permanently semiprotected because of this vandalism. Adding an abuse filter to catch this would mean that the article could be unprotected, which might lead to it being improved at a greater rate than now. Draftydoor (talk) 00:32, 27 November 2009 (UTC)

Okay, I'll bite. I tried unprotecting first. Wknight94 ^talk 04:47, 27 November 2009 (UTC)

I would be interested too

I had originally planned to ask for edit filter access a few weeks ago, but after working at the false positives page answering reports I came to believe that most of the false positives were triggering on the publically viewable filters, and that the remaining flaws in the edit filters were unavoidable and would simply have to be tolerated and cleaned up. But then I saw Triplestop's thread on AN and that he had gone on to change an existing filter to patch a hole that had apparently not been seen by anyone else up until now. And so I realize that false positives are not our only problem, and that there is plenty of room for improvement. I would like to help. I think I have a proven record of being cautious and am willing to answer questions as the other people requesting access have done.

If granted access, I would make changes that would eliminate false positives without letting through actual vandalism, and tweak existing filters to catch vandalism without triggering (excessive) false positives. For example, I think the changes I suggested above in the "party pooper" section would be a good idea because "party pooper" is not common in vandal edits and it wouldn't add much to the processing time of the filter since it would only appear in the same statement as the other words above. Other than the "party pooper" link given above, I don't have any urgent plans to make changes to the filters, although I will be constantly looking.

Having edit filter permissions would also help me find out the cause of problems that aren't apparent on first glance, if I'm understanding correctly that this test page I can't access is a place I can test out the filters (I know there's also test.wikipedia.org, but some of the cases that occur on enwiki can't be replicated there).

If you do not think you can use me for this position, please let me know why. -- Soap ^Talk/_{Contributions} 23:22, 26 November 2009 (UTC)

Do you know Regular Expressions? or anything else related to programming? Triplestop x3 03:15, 27 November 2009 (UTC)

Somewhat. I think I understand the code of all the public filters that I see, although some of them, such as Filter 9, are complex enough that I'd want to test any changes I make to make sure I'm not introducing a typo that would have unintended consequences. If I understand correctly there is a testing ground available to all approved users, and if not, there's test.wikipedia.org. I also have a text editor that supports syntax highlighting so I can more easily make sense of sequences like "(is|\'s|are|\'re|u\s*r\b)\s+(a|an|)\s*". (If I'm correct this is just a foolproof way to catch variants of "you're an..." There are some unfamiliar functions here, such as ccnorm, but they seem to be explained in the drop down menus. -- Soap ^Talk/_{Contributions} 04:22, 27 November 2009 (UTC)

Comment. Some time ago I noticed that Soap has been doing a great work on WP:Edit filter/False positives. Yesterday I came to a conclusion that he may need the EFM permission in order to handle false positives reports more effectively. To my surprise he applied for the EFM access before I even proposed him it to him myself. I studied the editing history of Soap and found nothing that can preclude me from grating him the EFM permission. Since there has been no objections for more than 24 h, I am going to do assign the EFM permission. Ruslik_Zero 16:47, 28 November 2009 (UTC)

You shouldn't have done that. Prodego ^talk 16:51, 28 November 2009 (UTC)

Why? Ruslik_Zero 17:03, 28 November 2009 (UTC)

Discussions on this page are slow, 24 hours isn't much time here. For example, from the 18th to the 25th (of this month) there were no comments on this page at all. Prodego ^talk 17:06, 28 November 2009 (UTC)

Comments appear very quickly if there is something to discuss. That there were nothing between 18 and 25 only indicates that nobody found a topic worthy of a comment. Beside I think it is reasonable to give this permission to the editor who has been answering all those false positives reports, when everybody (including you and me) seems to have forgotten that the page for FP reports exists. Ruslik_Zero 17:15, 28 November 2009 (UTC)

Thank you for granting me permission. I assure everyone that I won't do anything hasty. I don't see any urgent problems with the filters right now and so I have no plans for immediate action (Triplestop made my requested change to filter 46 yesterday). I plan to start out by working on existing filters, and only to fix obvious loopholes, eliminate conditions that I think would be uncontroversial (like the party pooper thing above), and fix typos (I remember once seeing a filter designed to trigger on users with <50 edits turn out to have been accidentally set to trigger only on >50 edits instead; it was fixed quickly but even in that short time there were lots of erroneous tags; in fact that's how I found the mistake). However, since I have heard some people complain about people getting edit filter access and doing nothing with it, I assure you that even if you don't see me making any changes to the filters right away, it only means I haven't found anything necessary and obvious to do. And I will continue to work at the False positives page, and although it's true that most of the reports on that page don't require anyone with edit filter access to respond to, I will appreciate having the ability to act for those rare cases that do. -- Soap ^Talk/_{Contributions} 19:02, 28 November 2009 (UTC)

I don't see a problem w/ granting EF permission to someone who already works in false positives. Likewise I don't see a reason to wait some unspecified period of time based on the local talk page churn. Protonk (talk) 21:31, 28 November 2009 (UTC)

Please explain

Can someone please explain how these made it through? They should have been stopped by filter 52 among others. Otherwise, how else to stop my jealous ex-girlfriend (who looks identical to Megan Fox, and loves pirates)? Wknight94 ^talk 04:52, 27 November 2009 (UTC)

More than 2% of action reach the condition limit now. So, it may not be surprising. Ruslik_Zero 14:03, 27 November 2009 (UTC)

They don't seem to match 52, so that would explain why it didn't match 52. My guess is it simply didn't match any filters, but I'm not going to check all of them. Prodego ^talk 15:56, 27 November 2009 (UTC)

Use of the confirmed usergroup to bypass filters

I have an idea to help users affected by false positives from disallowing edit filters. It involves changing all instances of the condition "autoconfirmed in user_groups" to "confirmed in user_groups". The condition would pass for both confirmed users and autoconfirmed, and would allow admins to control who can make certain edits. Any thoughts? Triplestop x3 02:55, 28 November 2009 (UTC)

Go ahead. Prodego ^talk 16:50, 28 November 2009 (UTC)

This makes sense. Ruslik_Zero 16:52, 28 November 2009 (UTC)

Date Linking

As we now have a bot de-linking and the community consensus was to not link dates (with notable exceptions) is it possible to have an EF to stop date linking, so that we can inform newer editors of the settled view on this activity? Or do we already have something and I'm not seeing it. Darrenhusted (talk) 21:11, 27 November 2009 (UTC)

Please see Special:abusefilter/2. Let me know what you think, though right now I'm not sure its worth it due to performance issues. Triplestop x3 21:30, 27 November 2009 (UTC)

Yes, but can it do the following (and I will plead ignorance if I my expectation of an EF are unrealistic); revert the action, then leave a "warning" (though nothing more than a level 1 notice) on the users talk page? Darrenhusted (talk) 21:45, 27 November 2009 (UTC)

PS: there was filter 106 which was triggered by delinking (and caught me a few times), though this was deleted in July. Darrenhusted (talk) 21:47, 27 November 2009 (UTC)

Well, the best it could do is display a warning message when they save the page, however I don't think it would be desirable to set it on disallow. See Wikipedia:Edit_filter#Actions_which_can_be_assigned_in_response_to_filtered_edits Triplestop x3 21:51, 27 November 2009 (UTC)

Then a warning before they save would have to do, it may make some reconsider adding wikilinked dates, and will help the delinkingbot further down the line and stop the task of date delinking from turning into an effort to akin to painting the Forth bridge. Is it possible to run a test for a few hours (again I'm not fully aware as to how this works, but a 24hr trial maybe to see the scale of the problem?). Darrenhusted (talk) 21:54, 27 November 2009 (UTC)

Yes, I am running the test now. Triplestop x3 22:00, 27 November 2009 (UTC)

Great, let me know how it goes. Darrenhusted (talk) 22:02, 27 November 2009 (UTC)

Please do not reuse older filters for newer things unless you created the filter, and know it will not be put back to its original use. For that reason, I redeleted Filter 2. Secondly, I strongly oppose using the abusefilter for anything as controversial as date delinking. Since 106 already existed for date delinking, please reuse that one. Prodego ^talk 22:11, 27 November 2009 (UTC)

This not for delinking, but rather to try to put a halt to wikilinking of dates, there is already a bot approved for delinking, and it is currently delinking January dates. Darrenhusted (talk) 22:45, 27 November 2009 (UTC)

I'm not sure it would be productive either to block an edit simply because the author linked a date, or to bother them with a warning to fix it. It is easy enough for another editor (or bot) to come delink the date, there is no reason we should block (or disrupt with a warning) good edits because of a linked date. Prodego ^talk 22:47, 27 November 2009 (UTC)

I only ask because the bot is currently going through and delinking almost every single wikilinked date, so to have new users relinking them seems a little counter productive. From what I can see most dates that are linked are down in new articles, and while new articles are tagged as such would it not be worth taking a look at a filter to tag date linking, even if there is no warning. Darrenhusted (talk) 22:58, 27 November 2009 (UTC)

Woops, sorry about that. I was not aware of the extensive Arbcom litigation over this. A bot is probably more suited to this purpose. Triplestop x3 22:16, 27 November 2009 (UTC)

I don't think an abusefilter is necessary either, but for the record, I don't think date linking is "controversial". There have been no issues since the ArbCom case, the community settled the issue peacefully, and we have had a bot delinking dates without complaint. Date delinking has become just another routine task. Dabomb87 (talk) 23:29, 27 November 2009 (UTC)

I agree with Prodego that blocking the edit just because a date happens to be linked isn't a good idea. Will just annoy editors who may link dates by force of habit. A bot coming along later and delinking it is no huge issue. –xeno^talk 18:37, 28 November 2009 (UTC)

I would say a logging filter might be useful. It is really a question of changing habits/culture - at the moment full-date unlinking bot is committed to visiting each article only once. I am scanning its contributions and picking up maybe 1 or 2 per 1000, mostly unusual formats and new stuff, but the odd reversion, intentional or otherwise. There will be more work to do to catch various date fragments, particularly easter eggs and just years will need a lot of community work and might not be (fully) bottable. Rich Farmbrough, 10:24, 30 November 2009 (UTC).

Auto-confirmed user de-autoconfirmed while vandal fighting

Main article: User_talk:Merlion444#Autoconfirmed_user.3F

Should we then make this user manually confirmed rather than simply restoring his auto status? Should we exempt Huggle users? Rich Farmbrough, 12:33, 30 November 2009 (UTC).

Restore the autoconfirmed status (as you have). That filter should not have been set up the way it was, I will talk to the person who is responsible about it. I've disabled that filter, and would note to everyone to not use the 'deautoconfirm' option under (pretty much any) circumstance. Prodego ^talk 18:53, 30 November 2009 (UTC)

Only tangentially related: does anyone know who to poke in order to get this presumably simple bugzilla fulfilled? –xeno^talk 19:07, 30 November 2009 (UTC)

Missing edits

The filter is missing more and more edits recently. I propose we start weeding out some of the tag-only filters in favor of some of the sneakier ones that are more difficult to catch. At least one long-term vandal follows a couple admins around blindly undoing their edits - that is far harder to catch then some of the obvious vandalism that is just getting tagged and ignored. Is anyone looking at the results of 23, 24, 39, 61, 117, 238, etc.? Thanks. Wknight94 ^talk 16:12, 13 November 2009 (UTC)

• I created filter 61, but I've been away for a while. The last time I checked, it worked fine, but it's almost impossible to check them all by hand. Could someone do it semi-automated? - Mgm|^(talk) 12:57, 2 December 2009 (UTC)

Quick question

I'm at a bit of a loss to understand why an edit like "nigger nigger nigger" would not be flagged by one or more filters? It seems like an obvious word to include in "bad words" filters. Delicious carbuncle (talk) 19:48, 30 November 2009 (UTC)

Hi, I'm working to extend some filters to user talk space where possible. I have already extended the "vandalism in all caps" filter (heres an example [25], and I'm seeing if the "Personal attacks by new user" filter can be extended safely. Triplestop x3 21:24, 30 November 2009 (UTC)

This is just a bit of a guess based on what I've seen in the past, but I think most filters aren't set to act in UT space because the condition limits would be too high. Or something. –xeno^talk 14:10, 1 December 2009 (UTC)

Most don't act in user talk space because it is not clear that swearing at other editors, etc, should be disallowed across the board. If established editors are permitted to do it, so are new editors, so that is why most of these filters are mainspace only. Prodego ^talk 21:56, 1 December 2009 (UTC)

Given the, uh, fungibility of the civility guidelines, that does make some sense. Perhaps having a separate set of filters for user talk pages (with somewhat laxer rules) might not be a terrible idea. I think there's also an argument to be made that IPs and unconfirmed editors should be subject to the same restrictions on user talk pages as they are on articles. This may not reduce the amount of user page vandalism, but it would result in less profane (and perhaps cleverer) user page vandalism. Is it worth discussing? Delicious carbuncle (talk) 22:12, 1 December 2009 (UTC)

I reverted filter 9 as there doesn't seem to be many hits in the user talk space. Filter 225 seems to be working well though. Triplestop x3 22:29, 1 December 2009 (UTC)

Request for permission

I would like to request permission to view and edit filters, although it is unlikely that I will edit them. I have an unblemished block log and am often pleasant to strangers in real life. Thanks. Delicious carbuncle (talk) 03:55, 26 November 2009 (UTC)

If it is unlikely you will edit them, why are you requesting the permission? Prodego ^talk 06:14, 26 November 2009 (UTC)

A fair question. I may wish to create or edit filters in the future and being able to review the changes made by others will enable me to become familiar with the use and content of the existing filters. If view-only permission were available, I would request that until I identified a need to edit, but such permissions do not yet appear to exist. Delicious carbuncle (talk) 12:58, 26 November 2009 (UTC)

Does this have anything to do with someone else recently being granted permission to edit the filter? Protonk (talk) 18:54, 26 November 2009 (UTC)

In that it made me aware of the possibility, yes. Delicious carbuncle (talk) 04:23, 27 November 2009 (UTC)

Not saying this is the case with DC, but I believe that some people want it for the sake of having it and being able to brag to their friends on IRC that they can see details of private filters not available to the general public. Being that security is only as strong as its' weakest link, I find the idea of giving abusefilter to non-admins a terrible idea. @Kate (talk) 09:52, 27 November 2009 (UTC)

Katerenka, I completely agree. I think that the entire idea that admins automatically have the ability to edit filters is misguided. There should be some evaluation of need and skills, ideally with the ability to view filters split from the ability to edit filters. There is an argument to be made that all admins should be able to view filters in order to respond to queries about disallowed edits, but I suspect that in practice the only people likely to look at filter criteria are those who have the ability to understand the regex and therefore the same people likely to want to edit the filters. Since it is possible for filters to block editors, I do not believe that any non-admins should be able to edit filter criteria. With that said, since we're handing out editfilter permissions to non-admins, I'm requesting it. If you check my permissions, you will see that I don't even have rollback, and I have no interest in participating in IRC. Delicious carbuncle (talk) 15:09, 27 November 2009 (UTC)

So you are saying this is request was made to try and make a point? Past community discussions have held that the edit filter may be granted to non-admins on a case-by-case basis. If you wish to revisit this discussion, then initiate a discussion: don't beat around the bush. –xeno^talk 15:18, 27 November 2009 (UTC)

Xeno, I don't see how you arrived at that conclusion. My request for editfilter permission is unrelated to my belief that non-admins should not be granted the permission. I restated that request in my most reply to Katerenka. Delicious carbuncle (talk) 15:29, 27 November 2009 (UTC)

I have not discussed abuse filters on IRC except with an admin in private. Triplestop x3 21:00, 27 November 2009 (UTC)

Is it actually possible for edit filters to automatically block editors? I thought that idea had been debated and discarded early on. Though it is possible it could be enabled in the future. -- Soap ^Talk/_{Contributions} 16:28, 27 November 2009 (UTC)

Not presently, no. Though, I suppose one could write a filter preventing a user from making any edit... –xeno^talk 16:37, 27 November 2009 (UTC)

Are you saying the blocking capability is not presently used in any filters or that it is not enabled for English WP? If it's the latter, then we need to edit Wikipedia:Edit filter to remove references to this capability. Can someone point me at the relevant discussions? Thanks. Delicious carbuncle (talk) 16:47, 27 November 2009 (UTC)

Correct. And fixed [26], thanks for pointing that out. –xeno^talk 16:50, 27 November 2009 (UTC)

I note that User:Soap's request has been granted, but mine is still pending, so I'm reiterating my request in case there was any confusion. Thanks. Delicious carbuncle (talk) 19:35, 4 December 2009 (UTC)

"I may wish to create or edit filters in the future and being able to review the changes made by others will enable me to become familiar with the use and content of the existing filters."

But you don't need the permission to review changes to the edit filter - simply click Special:AbuseFilter/history –xeno^talk 19:37, 4 December 2009 (UTC)

If you have an objection to granting me the edit filter privileges that I have requested, please express it. If you don't, would you mind granting the request? Thanks. Delicious carbuncle (talk) 19:47, 4 December 2009 (UTC)

I believe I have made my objection clear, the same objection I have with Chzz and Jakew, namely that if users are not using the rights for anything, they shouldn't have them. Since you seem to be indicating that you do not intend to use the right, I object to giving you them. Prodego ^talk 01:22, 5 December 2009 (UTC)

That's not at all what I said, however even if it were, that does not seem to be reason enough to refuse to grant me edit rights since both Chzz and Jakew were granted them and most admins have probably not edited filters let alone looked at the regex. Your comment when Jakew was granted edit filter permission was "Sounds good". (Incidentally, Chzz no longer has edit filter rights.) I have declared an intention to edit filters and asked for the appropriate permissions. I am hopeful that they will be forthcoming, but if there is a sincere objection please voice it clearly so that I may have a chance to fully address it. Thanks. Delicious carbuncle (talk) 04:22, 5 December 2009 (UTC)

My problem with Jakew is that he has still not used it. You said "it is unlikely that I will edit [the filters]". My belief is that those who don't want to use AFE should not have AFE. For admins who have granted it to themselves, there is nothing I can do. My objection is with the process, and my objection is that you say yourself it is unlikely you will do anything with the right. Prodego ^talk 04:42, 5 December 2009 (UTC)

I have already clarified that statement and re-declared my intention to edit filters. I am reluctant to assert that I will immediately begin editing filters because I think this is something that should not be rushed into hastily and because I have not confirmed support for new filters ideas or conditions. As you know from my earlier statements, I agree that the permissioning of filters is too lax, but I have been reminded that this is not the place for that discussion. Delicious carbuncle (talk) 05:17, 5 December 2009 (UTC)

I gave Soap the EFM permission, because he works on WP:FALSEPOS and needs it. I do not have any objections to you having the EFM access. On the other hand I am not going to do grant it to you myself, because I do not see a clear need for it. Ruslik_Zero 10:05, 5 December 2009 (UTC)

Would like to do one, need some help please

I've looked at current filters to try to get a feel for how to go about it, but I'm unsure. Essentially I want to test a filter that would at minimum tag and possibly disallow edits from IP users to a specific set of articles, eg Digimon Frontier, Digimon Adventure, The Filipino Channel (and a few others). Varying ranges of IP addresses are consistently inserting the same false text into these articles. The IP addresses all point to the same ISP in Indonesia. Rangeblocks have limited effectiveness, and it's been ongoing for close to 6 months. An edit filter would be amazing. So what would be the basic code to target IPs at those articles? I also have some keywords I could use, ie some of the stuff they are consistently reinserting. Cheers for any guidance or assistance. Also, note that once I know how to do this I will assist on other filters, thus your time won't be wasted showing me the basics. Also here's some diffs to show the type of changes they're doing to assist in devising a filter 1, 2, and 3. This isn't a request for a filter, it's a request to show me the basics or point me in the direction of really good guidance as I want to learn to do it myself. Nja²⁴⁷ 15:29, 3 December 2009 (UTC)

I would suggest WP:RFPP over an edit filter. Request semi-protection, as that seems to be what you want. Regards, MacMed^talk_stalk 19:09, 3 December 2009 (UTC)

If I wanted semi-protection I would have done it myself. Nja²⁴⁷ 19:35, 3 December 2009 (UTC)

Then you should go do it yourself. The abusefilter isn't to be used to semi-protect pages, that's what semi-protection is for. Prodego ^talk 19:39, 3 December 2009 (UTC)

I withdraw my request for assistance on how to get started. Apparently the regulars in this area aren't very helpful or friendly. Thanks anyhow. Nja²⁴⁷ 19:45, 3 December 2009 (UTC)

If you want to set up a filter for this, the first thing to do is to identify what the edits have in common, usually based on what the goal of them is. I'd be happy to help you with turning those traits into filter form. But, if you just want to stop IPs from editing a page, then you should use semi-protection. Which do you want to do here? Prodego ^talk 20:05, 3 December 2009 (UTC)

I am sorry if my first post wasn't clear. The edits have been done by a few specific ranges leading back to the same ISP. They hit they same articles with essentially the same vandalism, and it's been ongoing for some time. A rangeblock only results in a new set of IP's (again with the same ISP), and semi-ing several articles indefinitely due to one person's persistent vandalism wouldn't be a suitable solution. As the vandalism is the same each time, I assumed a filter could be devised. I've looked over the coding of some of the filters and could use guidance. I will provide any details you need, but I did want to try to get dirty myself as I would like to assist long-term, and don't want to put all the work on to others.

Anyhow as a starting point, if you consider the three diffs I provided in my initial post, one thing done in two of the articles was the addition of the same exact genres. Thus for those two articles (Digimon Frontier and Adventure), maybe the addition of those same genres would suffice in triggering. I'd have to devise other keyword triggers for each of the several articles targeted. Though I think this is enough as a start, plus I'm off for the evening, and I do appreciate your response and will get together any information you need. Cheers, NJA (t/c) 20:24, 3 December 2009 (UTC)

NOTE: The same genre additions (ie Survival and Horror) are common on a third article, see the diff. Thus the genre trigger could sort three Digimon articles the is targeted. NJA (t/c) 20:28, 3 December 2009 (UTC)

The vandal has also persistently added "Resident Evil" and "CBS Paramount Television" and/or "MGM," etc. to those articles. And the vandal has been IP range hopping as well (He used the ranges 202.70.5x.x, 114.59.x.x, 202.70.6x.146, and 118.137.x.x ranges, as well as several from the 125.16x.x.x range). - 上村七美 (Nanami-chan) | talkback | contribs 00:17, 4 December 2009 (UTC)

As you know, I know what he is vandalising. The point of this discussion is only to assist in code and to identify keywords to help devise a filter, not to report vandalism. NJA (t/c) 07:00, 4 December 2009 (UTC)

Hello, Nja247. Abusefilters are generally targeted to severe long term abusers or to disallow general behavioral patterns. If you must make a filter, the best way would probably be to block certain keywords to the pages mentioned. Or just disallow the whole ranges mentioned from editing the page. However, you have to keep collateral damage and performance in mind. Triplestop x3 21:53, 3 December 2009 (UTC)

Yes I realise this, though I wanted assistance with the coding. I identified three articles with specific keyword triggers above. I guess I'll have to take a stab at it later. NJA (t/c) 06:58, 4 December 2009 (UTC)

I will seek assistance independently with those who are accustomed to assisting others with code. Feel free to archive if this has taken too much space. Thanks anyhow and I apologise if this wasn't the best forum. If someone wants to help me get started then please do ping me on my talk page. NJA (t/c) 11:14, 4 December 2009 (UTC)

Suggestion

As a rollbacker and semi regular recent changes patroller, I often come across edits where some sort of profanity has been slipped into an article. This can get "lost" quite easily in the recent changes because, for example, adding "is gay" only adds 7 characters, compared with many on the list that will add/ remove several hundred or thousand. I wonder if a filter could be set up (or an existing one modified) to detect the addition of things like "is gay" or "sucks cock" etc since this seems a fairly common (and obvious) form of vandalism. It would also have the happy side effect of allowing quicker reversion of BLP vandalism. HJMitchell You rang? 12:02, 4 December 2009 (UTC)

Filter 9 already does that, but it has some exceptions intended to screen out false positives. If I'm reading the code right (I don't have my syntax highlighter here), it will not trigger if the user is autoconfirmed and editing in article space. This is probably because there is a legitimate need for people to write articles about gay people, and putting a blanket disallow would make that very difficult. -- Soap ^Talk/_{Contributions} 15:50, 4 December 2009 (UTC)

small grammar mistake, otherwise could go unnoticed but repeats itself in a VERY boring manner

well, start reading and check out all the "a edit"-s - i am bulgarian, but as much my modest education can reach, the correct is way to write/read it is "an edit" - always "an" when the following word starts with a vowel (like a,e,i,o,u, etc. :). Thank you for your attention. Best wishes to the reader and especially to this great project: Mike

PS. I am talking about the http://en.wikipedia.org/wiki/Wikipedia:Edit_filter page in case it was not quite clear. —Preceding unsigned comment added by 212.43.44.113 (talk • contribs) 11:11, 8 December 2009 (UTC)

Filter 276

I've created Filter 276 (logs) to handle this crazy proxy/spoofed IP insertion that is happening. See WP:ANI#Welcome to Wikipedldia!. Feel free to modify, turn on higher, turn off completely, and/or tell me I'm an idiot. Based on the number of IPs that we're seeing on it, I figured it was worth running as a filter without discussion. tedder (talk) 08:06, 21 December 2009 (UTC)

• • Incorporated into other existing filters and disabled. Thanks, NawlinWiki (talk) 17:54, 25 December 2009 (UTC)

User talk:NawlinWiki vandalism filter

If this private filter's name is an indication of its function, then I think it is an abuse of the right to edit filters. Sole Soul (talk) 04:14, 28 December 2009 (UTC)

Why would that be an abuse? tedder (talk) 06:55, 28 December 2009 (UTC)

This is a special case; given the amounst of vandalism his talk page gets, this is the most efficient way to minimize disruption. Someguy1221 (talk) 07:50, 28 December 2009 (UTC)

A recurring reason cited by filter editors to deny filter requests is that the abuse affects small number of articles (not a single talk page), and that the best way is to semi-protect because the filters cost a lot of resources. Also there is already a filter for some talk pages. I don't think this is the only talk page that gets vandalized a lot. Sole Soul (talk) 17:05, 28 December 2009 (UTC)

If I understand correctly, this filter only consumes 2 conditions of the 1000 condition limit, and the vast majority of all edits will bypass the filter after 1 condition, so it doesn't really consume very much resources. I admit I am uncomfortable with the idea of a user's talk page being protected with no unprotected talk page for IP's to post on, but the editnotice that this filter creates implies that NawlinWiki checks the filter log frequently to see if there any edits that were stopped by it, and will respond to the edits just as though they had been posted. -- Soap ^Talk/_{Contributions} 18:06, 28 December 2009 (UTC)

My main concern is the double standards, I don't think this would be allowed if a less privileged user requested a filter for his talk page (or for an article) with the same number of vandalisms. Although I assume that small things add up, so two conditions for a single page is still too much. Sole Soul (talk) 18:29, 28 December 2009 (UTC)

If you watchlist User talk:NawlinWiki, you'll may feel this is perfectly appropriate. A look through the history is probably sufficient, but make sure to check the history of individual pages listed at User talk:NawlinWiki/Archives.  Frank  |  talk  19:02, 28 December 2009 (UTC)

Sole Soul, if your page is vandalized 100 times a day, I'm sure you could be included in the filter if needed. Triplestop x3 19:07, 28 December 2009 (UTC)

I don't think so, and I don't think this page is the only talk page or article page that gets vandalized 100 times a day. Sole Soul (talk) 19:13, 28 December 2009 (UTC)

Nobody is suggesting it's the only one; what's being suggested is that this one is highly vandalized, and an edit filter is one tool in the arsenal to combat that.  Frank  |  talk  19:56, 28 December 2009 (UTC)

Nobody is suggesting that this one is not highly vandalized; what's being said is that since its not the only one, it should not be treated differently. Sole Soul (talk) 20:16, 28 December 2009 (UTC)

And indeed, the filter could be modified and then applied to others as well. We have to start somewhere.  Frank  |  talk  21:06, 28 December 2009 (UTC)

Indeed, we could, but if we don't you say it is not a problem, I say it is. Clearly, this was not intended as a "start somewhere" when the filter was created. Sole Soul (talk) 21:21, 28 December 2009 (UTC)

Looking at the log, It didn't catch anything in the last 4 days, and only 3 hits in December, 2 of them are false positives! Sole Soul (talk) 19:21, 28 December 2009 (UTC)

The filter is fine, though I'd ask NW to remember to disable it when it isn't needed. Prodego ^talk 20:54, 28 December 2009 (UTC)

The filter is not fine. "I'd ask NW", this shows the need to have an Edit Filter policy that is applied equally on everyone. Sole Soul (talk) 21:13, 28 December 2009 (UTC)

Yes. It is. Show me what policy this violates. Prodego ^talk 21:15, 28 December 2009 (UTC)

No, it isn't. You prove my point, there is no policy to prevent filter editors from creating anything, that doesn't make it fine. Sole Soul (talk) 21:27, 28 December 2009 (UTC)

Why should the edit filter be "applied equally on everyone?" User talk:NawlinWiki has more than 10x the number of watchers as User talk:Soap, and more than 5x the number of page views this month. (Ratios are similarly skewed for my own talk page in relation to NW's.)  Frank  |  talk  21:25, 28 December 2009 (UTC)

I meant to compare to User talk:Sole Soul: watchers (<30) and views (<100). Even wider disparity compared to User talk:NawlinWiki.  Frank  |  talk  21:31, 28 December 2009 (UTC)

It's obvious that that is not what I meant when I said "an Edit Filter policy applied equally on everyone". Sole Soul (talk) 21:33, 28 December 2009 (UTC)

It's not obvious, actually, I have no idea what else you could have meant. Try being a little more helpful? ╟─TreasuryTag►UK EYES ONLY─╢ 22:20, 28 December 2009 (UTC)

So you disagree with this statement: "an Edit Filter policy [should be] applied equally on everyone", keeping in mind that I said above "...with the same number of vandalisms". Sole Soul (talk) 22:50, 28 December 2009 (UTC)

Where did I say I disagreed? (Not a rhetorical question, please provide a diff.) I said, "I have no idea" what you meant. ╟─TreasuryTag►without portfolio─╢ 23:03, 28 December 2009 (UTC)

That was a question, I forget to put the question mark. Sole Soul (talk) 23:10, 28 December 2009 (UTC)

OK; what words that I typed made you think that I disagreed? What prompted you to ask the question? (Not rhetorical questions, please provide the answers.) All that I said was, "I have no idea" what you meant. ╟─TreasuryTag►senator─╢ 23:12, 28 December 2009 (UTC)

I have no interest in playing games with you, I'm reading all responses here so no need to send messages to my talk page. Sole Soul (talk) 23:21, 28 December 2009 (UTC)

I see, so you're just going to falsely accuse me of making derogotary remarks and then decline to engage in the follow-up discussion? I'm not impressed. ╟─TreasuryTag►assemblyman─╢ 23:23, 28 December 2009 (UTC)

There has been no policy, and the problem with making one is that people need to be able to quickly implement (some) filters in response to ongoing (serious) vandalism. If there is ongoing serious vandalism to a single page, any page, it is perfectly fine to create a filter to stop it. Filters that aren't the most useful ultimately get disabled by other abuse filter editors (often me). Prodego ^talk 22:09, 28 December 2009 (UTC)

(Note: I appreciate it when I saw you turning some filters public when secrecy is not needed) So you think that the lack of policy is for a specific reason, and you take advantage of the lack of policy in a discussion that has nothing to do with that specific reason? Sole Soul (talk) 22:32, 28 December 2009 (UTC)

I'm explaining why I believe a policy like the one you are suggesting be implemented here would not be feasible. Prodego ^talk 22:35, 28 December 2009 (UTC)

I'm talking about your saying "show me a policy". Sole Soul (talk) 22:39, 28 December 2009 (UTC)

(e/c)This is exactly the reason we don't have such a policy. If such a policy would prevent creating an abuse filter to stop the kind of vandalism that hits NW's talk page, it would be putting bureaucratic nonsense over actually using the filter for what its designed for. Mr.Z-man 22:40, 28 December 2009 (UTC)

The policy would not prevent such a filter if the community is fine with a filter for a single page. The policy would ensure that the standards is equally applied in all similar situations. Sole Soul (talk) 22:57, 28 December 2009 (UTC)

What standards? Prodego ^talk 23:07, 28 December 2009 (UTC)

In this case a page with high number of vandalism. Sole Soul (talk) 23:16, 28 December 2009 (UTC)

Until User:JarlaxleArtemis and his army of idiots at 4chan /b/ grow up, I would like to keep this filter. If and when a consensus is reached that I must rely on semiprotection instead, I'll do that. In the meantime, I am going to be away for about a week, and I would appreciate the filter being left on during that time. NawlinWiki (talk) 23:09, 28 December 2009 (UTC)

I don't know, for the record, I think this filter is fine. There doesn't need to be some massive beurocratic mechanism like botapproval or something like that for these things. Theres a patently obvious need here, the filter is the least intrusive means to stop this vandalism, and it works. Well within the spirit of everything at Wikipedia. If it works, go with it. --Jayron32 05:44, 29 December 2009 (UTC)

Permission

I would like to request the edit filter right. I would like to create filters that will block any malicious-like scripts. I would also enjoy being able to view private filters. Btilm_{Happy Holidays!} 05:20, 24 December 2009 (UTC)

I can assure you that there is nothing enjoyable in the codes of private filters. Ruslik_Zero 20:12, 24 December 2009 (UTC)

Well, I will at least like to create a filter for malicious scripts. Btilm_{Happy Holidays!} 01:04, 25 December 2009 (UTC)

Can you provide an example of such malicious scripts and what code you would write to block it? Triplestop x3 01:09, 25 December 2009 (UTC)

There are many little codes that could possibly freeze one's browser. That would be this code's only harm. Here is what a vandal would type, with or without the <script> and </script>. <script>while (1==1) document.write('hello')</script>

This is a script that keeps writing hello until you close the frozen browser.

This are the conditions: !("sysop" in user_groups) & (ccnorm(added_lines) rlike "(<script>)?(\n)?while \(?.*==.*\)?\{?(\n)?\{?(document.write|alert)\('.*')\}?(\n)?\}?(</?script>)?")

I would first test the code by flagging the edit in the edit filter log. Once I get it right, I would disallow the code. Btilm_{Happy Holidays!} 17:36, 25 December 2009 (UTC)

As far as I know, the MediaWiki markup doesn't allow scripts, so what you describe wouldn't happen on Wikipedia... ╟─TreasuryTag►consulate─╢ 17:38, 25 December 2009 (UTC)

I am talking about if a vandal puts it on there js page, or if a user is testing something out. Btilm_{Happy Holidays!} 17:52, 25 December 2009 (UTC)

Is "vandals injecting malicious code" a recurring problem? Is there even a single precedent of vandals doing js mischief? What a vandal puts in his own .js will affect only his own browser anyway, not others. Looks like a solution in search of problem to me.--Zvn (talk) 20:26, 25 December 2009 (UTC)

If you know of some way to run scripts for people other than yourself that doesn't require admin rights, please contact securitywikimedia.org, as that would be a rather serious issue that should be fixed in the software. Mr.Z-man 19:04, 26 December 2009 (UTC)

You're kinda right, Zvn. Maybe an admin should try that instead of a block sometime. Another reason, now that I think about it, is to be able to create new filters wanted here. One is the addition of "you." I could do that. I will give an example.

Conditions: !("sysop" in user_groups) & (article_namespace == 0) & (lcase(added_lines) rlike "\b(you('?re?(el(f|ve))?s?|'?ll|)|y'?all'?s?)\b")

I would test it by flagging it; and then, once I get it going, I would then choose to flag the edit and tag the edit (the tag would be something like possible addition of the word you). Btilm_{Happy Holidays!} 02:49, 27 December 2009 (UTC)

Can you elaborate on what types of edits there should or should not be an abuse filter for? Triplestop x3 04:13, 27 December 2009 (UTC)

Well, the first thing I would like to point out is, if I get the abusefilter right, I would never check the box to revoke the autoconfirmed status, because I can't think of a single instance to use it on; and due to the fact that probably every filter has, or at one point had, a false positive. For example, a filter that detects when a user writes a phone number wouldn't be the best idea in the world. There could be many false positives, where phone numbers should be included in an article, like 555 (telephone number), 867-5309/Jenny, and Telephone number. A filter that wouldn't be so bad would be one that detects when new users create an article via the article wizard and leaves, under external links, example.com. Articles that have that are usually deleted. The criteria for that would probably be something like this: edit count - under 30 edits and account age - 2 days or less. If I do get the right, it would help me answer the entire why did X get flagged, but not Y business on private filters. This process would also help me discover false positives for private filters, and I will still be looking for false positives on public filters. I would probably fix some false positives and moniter the requested filters. Btilm_{Happy Holidays!} 05:32, 27 December 2009 (UTC)

 Not done I can't say with confidence after reading the above discussion that you have the knowledge of what the abuse filter is to be used for, nor can I say that you know how to code for it. NW (Talk) 16:43, 29 December 2009 (UTC)

Well, please give me a clear case to code and I will code it to the best of my abilities. Btilm 17:42, 29 December 2009 (UTC)

Please? I really have a chance of doing it. I do understand it, but you don't have confidence in me. So how about some Q and A? Btilm 03:01, 1 January 2010 (UTC)

Proposal

A proposal has been made here which may be of interest. ╟─TreasuryTag►assemblyman─╢ 15:00, 29 December 2009 (UTC)

Filter 68 turning up a lot of false positives

Filter 68 is denying a lot of legitimate moves. I notice that this is because it's been recently made much stricter, and I would just like to bring that to people's attention, and request that if we think it should stay strict, perhaps a message of some kind could be written that explains clearly why the edit is being denied so that people will be less frustrated. Or, is teh message deliberately left out because this filter is private and giving more details would spoil the purpose of it? Also, I wonder if filter 10 could be scaled back a bit because it seems to be catching more falsepos than real ones and most of the edits that would trigger it legitimately would be caught by various other filters. -- Soap ^Talk/_{Contributions} 21:12, 1 January 2010 (UTC)

Limits on 68 drastically increased. Prodego ^talk 23:57, 1 January 2010 (UTC)

Proposal: Make it easy to undo false positives

I propose that the filter log contain an Undo link or button for every entry. Then experienced editors can easily undo false positives, allowing the user's edit to be performed normally. Otherwise, there appears to be no way to undo false positives. David spector (talk) 01:57, 5 January 2010 (UTC)

That wouldn't work if there were any edits to the page after the one you are trying to 'redo'. It is basically the opposite of the 'undo' feature, with the same limitations. Of course this is only the case for filters set to disallow. Prodego ^talk 03:11, 5 January 2010 (UTC)

To be honest, false positives on "disallow" filters indicate either a problem with the filter or the post. If it's not a problem with the post, then the filter should be corrected. I fear that if something like this were implemented, it would reduce the number of false positives reported and problems would potentially go more unnoticed. I agree it would make things easier for editors, but I think it would also be a net loss, as confusing as that sounds. --Shirik (Questions or Comments?) 12:17, 5 January 2010 (UTC)

I think he wants an Undo button for the people who watch WP:FALSEPOS, not the people who trigger the edits themselves, which is what it looks like youre thinking. And since every edit that triggers a filter preserves the edit in the log, it is possible for FALSEPOS watchers to go ahead and complete the edit, but only by pasting the code in manually. I agree an Undo button would be nice, though I wouldn't call it Undo personally, as that implies the edit went through and was reverted, when it never went through at all. -- Soap ^Talk/_{Contributions} 13:49, 5 January 2010 (UTC)

I assumed he meant allowing all editors (maybe all autoconfirmed) to do it; naturally, the person who triggered the filter wouldn't be able to override it, but I viewed the button as a "second opinion" that would let it go through. If it is more restricted as you suggest, then my concerns have less weight, but if it's available to all autoconfirmed users, I feel that it would end up being abused (in good faith) more than it would be used. --Shirik (Questions or Comments?) 14:10, 5 January 2010 (UTC)

As a side note, perhaps an interesting possibility would be to make this available in the same place as false positives are reported, this way you don't have the conflict of "need for report" vs "ease of use". You get both. At first glance, that doesn't seem technically possible, though. (At least not without major changes in how things operate.) --Shirik (Questions or Comments?) 14:12, 5 January 2010 (UTC)

I implemented a proof-of-concept for the idea which is functional with the current implementation of the edit filter and Wikipedia. What could be done is a (retry) link could be added to the {{falsepositive}} template which links to an external tool which researches the false positive, pulls up the wikitext, and redirects the user to a preview containing the edit as was originally tried. Here's an example link on the reported "Sophie Wilson" edit: retry. Please note that it's a proof-of-concept and, thus, is a little rough around the edges in presentation. But it's functional (at least on Chrome - please let me know if other browsers are having issues). Feel free to click on it; it only takes you to a preview, it will not (and cannot) do the edit on its own. --Shirik (Questions or Comments?) 23:25, 5 January 2010 (UTC)

Note: It would appear Internet Explorer is complaining about "possible cross-site scripting" when using this tool. I'm not sure what it's picking up and changing, but it still seems to work. --Shirik (Questions or Comments?) 23:41, 5 January 2010 (UTC)

Admins can now see private filters without abuse filter editor rights

The permission has been enabled in T22721. This implies that there is no longer a need to (self-)assign the abuse filter editor permission to admins to be able to check private filters, so we can consider removing the permission from admins who don't edit filters, or other ways to assign it, as some had suggested. Cenarium (talk) 18:54, 26 December 2009 (UTC)

Could we consider making the whole abusefilter bundle part of the sysop package? Admins who want to edit the filters (or think they might want to in some never-to-be-realized future) self granting the right just creates clutter in my opinion. I would keep the abusefilter right around for those who have it but aren't sysops, but allow only 'crats to manipulate it. ⇌ Jake Wartenberg 19:20, 26 December 2009 (UTC)

I don't think being an admin should automatically allow you to edit the abuse filter. We choose admins for good judgment, not good programming skills. Being a good admin does not really prepare you for editing these filters as it requires technical ability beyond what is expected by an admin. While I don't see any harm in all admins being able to see the filters, I think being an admin and being able to edit the filters should remain distinct positions. Chillum ^{(Need help? Ask me)} 20:24, 26 December 2009 (UTC)

We allow admins access to other technical parts of the site -- the abilities to edit complex, highly used templates, the mediawiki namespace, and the spam and title blacklists are all part of the sysop package, among other things. We trust admins not to use their tools in places where they don't have the necessary expertise. I also don't see how this creates a "separate position" when any admin may grant herself the right on a whim. ⇌ Jake Wartenberg 02:13, 27 December 2009 (UTC)

The abusefilter differs from all of your examples as it's a tool that can make a "mark of Cain" entry on the permanent record attached to an editor. –Whitehorse1 22:35, 27 December 2009 (UTC)

This is true; however, I think it should be more widely known that the edit filter (renamed from its original name of abusefilter) carries a lot of false positives, and even the best editors get tagged by it once in awhile. For example, here's an administrator with 15 of them, none of which are actually indicative of "bad faith" edits. Hopefully more people will come to think of it as a mostly benign way of categorizing people's edits rather than a permanent reminder of past vandalism. -- Soap ^Talk/_{Contributions} 03:10, 28 December 2009 (UTC)

I would support cleansing of the userright from admins who do not edit the filter to make it easier to find admins who do. The suggestion by Jake is a separate issue and needs its own discussion as each time we discussed it (see archives), the status quo was upheld for various reasons including the fact that an admin having to grant themself the userright is a bit of a "speed bump" for them to go over - a deliberate act that hopefully makes them realize that it is a serious tool that requires caution in its application. –xeno^talk 15:28, 29 December 2009 (UTC)

Remove flag from admins who don't use it

So, a proposal, now that circumstances have changed considerably: Remove the flag from those that don't use it. The admin can always grant it back, so this shouldn't be a big deal. Perhaps we could loosely define inactivity as no actions withing a three month period? ⇌ Jake Wartenberg 22:19, 30 December 2009 (UTC)

I'd so go for it. Make the list useful again. –xeno^talk 18:36, 31 December 2009 (UTC)

I just made an initial run, disabling the userright for about 130 admins who had never edited the filter [27]. I skipped anyone who had added it recently (within the last 3-4 months). –xeno^talk 16:17, 6 January 2010 (UTC)

Thanks for doing that, Xeno. How did you find the list of admins who have edited? Curious because you removed mine even though I've edited a filter in the last month. No big deal, of course. tedder (talk) 16:24, 6 January 2010 (UTC)

Hm, that's strange. Your name did not appear in Special:AbuseFilter/history; which is where I generated the list of those who did edit. –xeno^talk 16:33, 6 January 2010 (UTC)

Look on 2 Jan 2010. In any case, it isn't a big deal. I'm perfectly capable of readding myself, which is what I did (your edit summary was clear in that). tedder (talk) 16:37, 6 January 2010 (UTC)

I'd recommend using this instead; it's more compact and it's easier to load up 500 at a time instead of 50. (It doesn't give you the details of each edit though.) -- Soap ^Talk/_{Contributions} 16:38, 6 January 2010 (UTC)

Thanks, Soap. tedder, I don't know why you aren't showing up in the history link - my apologies. Will use the better link and rv self on any I inadvertantly removed. –xeno^talk 16:43, 6 January 2010 (UTC)

I see myself on both links- dunno why not on yours, but it's no big deal. You left a good edit summary and those affected are admins who can resolve it very easily. tedder (talk) 16:50, 6 January 2010 (UTC)

Are you quite sure it's not a big deal? ;) ╟─TreasuryTag►First Secretary of State─╢ 16:52, 6 January 2010 (UTC)

Must be some funky filtering on my end. Looks like changes for private filters aren't shown to admins without the EFM right in the history. I've filed a bugzilla on that. Re-added to a handful of users - and as you say, this is just a housekeeping thing so there's no problem at all if an admin wants to re-add it. –xeno^talk 16:56, 6 January 2010 (UTC)

FWIW and to avoid further confusion, the log only goes back to 3 July 2009, so it's missing a couple months that can only be found in the abuse filter history. Amalthea 13:22, 7 January 2010 (UTC)

Yep, I noticed that. Which is why (after I figured out why special:AF/history was hiding things from me) I extracted a complete list from the history going back to the birth of the filter in order to revert the handful of erroneous removals from folks who had only edited private filters. –xeno^talk 14:30, 7 January 2010 (UTC)

Proposal: tagging removal or change of images in articles

Edits by non-autoconfirmed editors should be tagged that either remove an image from an article or add an image to an article. MediaWiki:Bad image list has significantly cut down vandalism from inappropriate images, but that depends on a particular image already having been identified as problematic. Removal of an image generally by new or anonymous users is likely to be blanking vandalism or misguided censorship of a controversial topic, and the existing blanking tags would not catch this unless the user deletes a larger portion of the text along with the image. If the image is a validly used non-free image, it can get automatically tagged for deletion as an orphan if no one catches its removal in time, which at minimum causes wasted labor. postdlf (talk) 17:39, 5 January 2010 (UTC)

Note: I have posted this request at Wikipedia:EF/R#New_user_removing_images --Shirik (Questions or Comments?) 03:42, 6 January 2010 (UTC)

There was filter 131, which prevented users from removing controversial images. Perhaps, it can reused. Ruslik_Zero 17:02, 6 January 2010 (UTC)

I went a different route and created Filter 280, primarily because Filter 131 seemed to be directed at specific images. It currently passes my unit tests, but I haven't activated it in case anyone else wants to take a look at it for performance reasons (the regex is a little long for my liking). If someone with more experience than I could take a look at it and say it looks OK, I'd like to put it into log-only mode to see what it's picking up and how long it takes. --Shirik (Questions or Comments?) 21:56, 6 January 2010 (UTC)

Documentation

Is there another source of documentation for the rules format other than this? It seems to not document many functions I see being used, like any_content, str_replace, and rcount. I'm sure there are others.

I'm also curious for more details about how it handles regexes. Do paranthesis need to be escaped, for example, and can I include newlines, is it multiline, etc. Shadowjams (talk) 08:32, 7 January 2010 (UTC)

It uses perl-compatible regular expressions (PCRE). Anchors do not match newlines. Beyond that I can't say much; PCRE are a very complex beast. --Shirik (Questions or Comments?) 09:47, 7 January 2010 (UTC)

Yes, you can use newlines though I think we prefer the escape character \n rather than actual newlines. The start and end characters match only at the very end (i.e. single line), and yes parenthesis should be escaped. The functions have brief descriptions in the drop-down menu when editing filters, which helps complement the incomplete documentation. Dragons flight (talk) 10:20, 7 January 2010 (UTC)

Category redlinks

I have no idea if this is technically feasible or not, so I thought I'd bring it up for discussion.

Regularly, people try to add non-existent categories to pages, often treating them as a pretty basic tagging system. A good number of these are good-faith edits, and a good number are vandalism - see, eg, this for an example of the latter.

This is an unusual problem, though, because either way it needs intervention. If I try to add Category:Fascist pigs to an article, it needs to be removed, but if I try to add Category:Fascist organisations in Belgium, it still needs a user to correct it to Category:Belgian fascist parties. As matters stand, you can go to a redlink and find any page trying to be a member of that category, but you can't easily find a collection of articles with category redlinks.

So, it seems a possible use for the edit filter. This boils down to two questions:

a) is it possible to mark an edit which adds an invalid category? I'm guessing the expression would need to be something along the lines of (adds category link) and (linked page in category namespace does not exist), but I don't know if that second part is feasible.

b) ...and would we want to? This seems like something which would need tagging, and potentially could produce an immense amount of results, but on the other hand, it would allow us to patch up a decent-sized hole in the category system.

Thoughts appreciated. Shimgray | talk | 12:28, 7 January 2010 (UTC)

Request for Edit Filter Manager Permissions

I want to start off by saying that I recognize this will likely be a highly controversial request. I will take no offense if it is denied outright. However, I figured that there is no harm in making the request, but potential gain if it is granted, so either way there's no net loss. I'd like to request the edit filter manager permissions. I recognize that I'm not an administrator and that granting this access is rare, but I have two primary reasons for the request:

• I feel that I can help out with outstanding requests for filters. I am a software engineer that works on safety-critical systems, so I feel that I not only have the expertise to do this, but also I understand the necessity of getting it right the first time. If I were to create a filter, I would not only test it thoroughly, but I would also not activate it until it has had at least one second opinion, preferably from someone with more experience than myself.
• I also am currently working on a bot, part of CollabRC which uses artificial intelligence to judge vandalism and either raise alerts or outright revert. I feel that the ability to view some of the private filters may give me insight on certain "genes", such as those I have already developed, that would be appropriate to build to benefit the bot.

I have a history of anti-vandalism efforts and I hope this is sufficient to acquire your confidence that I would not misuse the tool, but I totally understand that my lack of history in this particular area is lacking, so I would also not take offense to a denial of this permission. I would note that, should the permission be granted conditionally (such as on the condition that I never activate a filter on my own) I would certainly adhere to those conditions. I tried to be as brief as I could while explaining the full reasons behind my request; if you have any questions for me please don't hesitate to ask. I will be more than happy to explain in more detail any questions about my qualifications, concerns about my integrity, or details on my reasons for desiring this permission. Thanks. --Shirik (Questions or Comments?) 05:37, 1 January 2010 (UTC)

• Hi, Shirik. I believe you may be very qualified for the bit. Would you apply your methods ("Genetics") to abuse filters? If so, how? Triplestop x3 06:26, 1 January 2010 (UTC)
  • If I could, I would. (By the way, before we go too far, I want to clarify that I didn't "invent" these methods -- it's a well known artificial intelligence paradigm; there are even two articles on it: genetic programming and genetic algorithms, but I digress). There's two sides to this. Firstly, the concept of genetic programming relies on the ability for a software concept (in the case of an edit filter application, it would be more like a regular expression, though that's not entirely how WP:EF works) to adapt on its own. Unfortunately, edit filters, as they are designed now, cannot do that. However, as the bot continues to train itself, I would be very interested in taking those genes that it determines to be extremely valuable and turning them into edit filters. Naturally, this would still have to go through intense scrutiny because, in this case, not only is it something we don't want to apply indiscriminately, it is something that wasn't even developed by a human, so it would need careful review. Should User:CollabRCBot happen to determine some interesting components of inappropriate content that are representable by an edit filter, I would be more than willing to try to adapt it to one. Unfortunately, because the very nature of genetic algorithms are based off randomness and survival of the fittest, I cannot really predict if what it produces will be able to be an edit filter. Sorry for such a detailed response. --Shirik (Questions or Comments?) 06:34, 1 January 2010 (UTC)
    • We do have many filters that work the way you describe, however none of them really go by that point system. Can you discuss how you would handle issues such as accuracy, false positives and performance? Triplestop x3 20:25, 1 January 2010 (UTC)
      • Certainly. I think one of the best ways to describe my efforts at ensuring accuracy are through examples of past work. One of the things I to ensure accuracy is to write unit tests for operations, such as those found here. What I would do is, before implementing the filter, I would create a set of test posts which include (1) things that should be caught by the filter and (2) things that should not be caught by the filter (but preferably are very close). Then, after implementation, I would test the filter against these test posts. The chances of writing both the test and the filter wrong are much smaller than just the filter itself. Additionally, I strongly believe in code review and would still look for a review to ensure accuracy. Addressing performance, I realize that string manipulation is much more complex than others expect it to be. Certain things should not be filters as they are better addressed by users on huggle, etc., especially if the filter will affect a very small proportion of articles but would still have to test every edit. I would plan to incorporate short-circuit logic in an effort to minimize the number of operations that must be done, and leave string manipulation for the very end of the filter. The more we can eliminate early in the filter test (such as restricting to only non-autoconfirmed users), the better. Finally, with regards to false positives, I would hope that it never happens, but perhaps that's too utopian. If I were to create a filter, I would be monitoring it significantly for at least the next week or more (depending on how complex it is), watching what trips it. If there are any false positives, then that indicates a problem to me - the next question is "is the fix worth it". False positives, in my opinion, are unacceptable except for those that are tags (and even then, false positives should be as rare as possible), so those filters that disallow a post, etc., must be fixed or disabled. If the performance penalty for fixing the filter is too significant, then it should be disabled and left for more complex methods like human vandal-fighters or bots. Unfortunately, this is a very case-specific judgement call and is a bit difficult to discuss abstractly, but I hope that I explained my thought process well enough for you to understand my plan of attack. --Shirik (Questions or Comments?) 21:39, 1 January 2010 (UTC)

Ok with me. Prodego ^talk 23:54, 1 January 2010 (UTC)

 Done Prodego ^talk 19:59, 8 January 2010 (UTC)

18

Why was this filter turned off? This is a important issue, given that there are over 3500 results for 'insert non-formatted text'.— Dædαlus ^Contribs 00:13, 12 January 2010 (UTC)

I was actually meaning to bring this up myself, after seeing the hundreds, possibly thousands, of edits User:TTTSNB made to remove them a few days ago, but I forgot. I suspect it was turned off because a lot of times people will do things like that and insert legitimate content in the same edit, and even though the filter was not set to Disallow, the people who make test edits are likely new to Wikipedia and thus easily confused, and may not know how to remove the "test" part of their edit and thus get discouraged. But that's just a guess. -- Soap ^Talk/_{Contributions} 00:18, 12 January 2010 (UTC)

My (limited) understanding of the reasoning behind turning it off was exactly as Soap described: that it is too WP:BITEy and discourages new users who are getting their first edits blocked right away. As I mentioned to User:TTTSNB, this should be a pretty trivial thing for a bot to clean up, and it is my understanding that he has already filed a bot request for it. --Shirik (Questions or Comments?) 06:46, 12 January 2010 (UTC)

I suppose it could be revived as a tag-only filter, though on WP:FALSEPOS I can see that there are people who feel aggravated even by tag only filters. -- Soap ^Talk/_{Contributions} 01:24, 15 January 2010 (UTC)

Filter 82

I'm sure wikipedia has rules against posting 4shared download links to illegally uploaded media. Filter 82 was enabled to log, but not disallow. I propose that we edit it to only disallow 4shared links, then set it to disallow in order to take care of that problem.— Dædαlus ^Contribs 02:47, 28 December 2009 (UTC)

Note: this user contacted me on IRC earlier and I suggested they bring this to the wiki because there are issues here I don't feel confident handling by myself.

The filter recently logged this edit which was only inserting a malformed ref tag ... I've looked at the code and am not sure what is causing it. I am worried that there may be a quirk in MediaWiki that causes some edits to trigger the filter even though they don't contain a [nowiki] tag. I also want to point out that I dont think setting this filter to work only on 4shared would be a good idea because there are other hosting sites, and all of them can be reached via TinyURL and other URL forwarders which would not be stopped or even logged if this filter were set to trigger specifically on 4shared. Perhaps "4shared" could be added to a blacklist contained on one of the filters that is already in Disallow mode, and this one kept as log-only to keep track of less-blatant edits and possible false positives. I am sorry if I sound negative, because I agree that this filter needs to be improved. -- Soap ^Talk/_{Contributions} 03:03, 28 December 2009 (UTC)

Note that TinyURL and other such services are generally blocked on the global spam blacklist. If we want to block such links, we should also use the spam blacklist, as it can probably do it more efficiently. Mr.Z-man 04:00, 28 December 2009 (UTC)

Can the spam blacklist be configured in such a way that it will catch links surrounded by <nowiki> tags? I thought that the reason this filter was needed was because it couldn't. -- Soap ^Talk/_{Contributions} 04:03, 28 December 2009 (UTC)

The the point of the filter is to catch any link in nowikis; if we want to specifically block 4shared links, that should be done with the spam blacklist. Mr.Z-man 04:56, 28 December 2009 (UTC)

This is a reply to Soap- apparently the nowiki tags doesn't cause tinyurl to be blacklisted. Interesting. http://tinyurl.com/foo tedder (talk) 06:54, 28 December 2009 (UTC)

So, can we get some headway on this? If tinyurl is blacklisted, then I see no problem with adding it to the filter to block 4shared and tinyurl.— Dædαlus ^Contribs 05:06, 31 December 2009 (UTC)

Oh, Tedder added it a few days ago. Sorry, I'd thought you'd heard about it. I'll be watching it for false positives because I still dont really know what caused the [ref] bug. (And it is still turning up false positives such as this, although that particular user was able to fix it by removing the unnecessary nowiki tags from his edit.) -- Soap ^Talk/_{Contributions} 05:17, 31 December 2009 (UTC) I know that link will just give an error, but all that it really shows is that he's turning "disallow" on.)

(ec)Huh, weird timing. Yeah, I turned it on- see also this thread. tedder (talk) 05:20, 31 December 2009 (UTC)

Soap- that diff is definitely a FP, but .. how concerned should we be? It's not something that should be done, even if it's just a technical issue, not a bad-faith edit. tedder (talk) 05:29, 31 December 2009 (UTC)

Tedder, I'll be leaving a similar note on your talk page if you know how to edit these things, but, per this edit, it needs to be adjusted so that 4sh@red isn't allowed either.— Dædαlus ^Contribs 05:00, 1 January 2010 (UTC)

Modified filter slightly. tedder (talk) 02:31, 2 January 2010 (UTC)

I need some help here. It turns out the mod to the filter doesn't catch elements like it should. I experimented, and I don't understand why it isn't working. Can an experienced editfilterer look at it and see if you can assist? Thanks. tedder (talk) 04:57, 2 January 2010 (UTC)

Daedalus' edit went through because he has a high enough edit count (by a very wide margin) to escape the condition set by the filter. -- Soap ^Talk/_{Contributions} 00:02, 5 January 2010 (UTC)

Needs to be adjusted again, per this edit. Com'on guys, we're obviously putting a dent in what he's able to do, soon, it won't be anything.— Dædαlus ^Contribs 23:41, 4 January 2010 (UTC)

A new approach in Filter 278 may be able to solve this problem; as I've said before I'm not really confident that Filter 82 is ever going to be able to. Given that it was you who brought this problem to our attention, I assume it would be OK to paste the code of the filter to you private in an email or on IRC but just to make sure there isn't some firm rule against it I will wait for someone else to tell me for sure. -- Soap ^Talk/_{Contributions} 01:03, 5 January 2010 (UTC)

That should help, yes, though it doesn't invalidate filter 82 for non-4shared links. And it doesn't catch variants well- for instance, "4sh@red", which we've seen coming through. I don't know what the rule is about sharing the filters, nor do I think it matters- in other words, feel free to share with Daed. We didn't think of the autofilter/number of edits restriction- duh! tedder (talk) 01:06, 5 January 2010 (UTC)

278

I am liking what I am seeing, this may be just what we need.— Dædαlus ^Contribs 01:11, 5 January 2010 (UTC)

I see about 1 false positive, and that was adding a bunch of spam links, in violation of wikipedia is not a collection of links. That in mind, we have gotten almost nothing but positive hits. I think it's time to disallow.— Dædαlus ^Contribs 05:18, 15 January 2010 (UTC)

I made a slight change to 278 today to expand its capabilities slightly; I think it should be given a little more time just to verify nothing's wrong with it, but it's just a slight expansion so I'm not expecting any problems. --Shirik (Questions or Comments?) 10:41, 15 January 2010 (UTC)

 Done Reviewed independently and flipped on disallow. --Shirik (Questions or Comments?) 10:04, 17 January 2010 (UTC)

Requesting EFM permission

Hello,

The devs enabled AbuseFilter on Turkish Wikimedia two days ago, and now we are in need of already developed filters you guys have over here.

I need to access the hidden filters and to be able to export them. I am an admin and crat on Turkish Wikipedia (click to verify), the founder of Wikimedia Türkiye, and has been a Wikimedian for over 3 years, so I can be trusted.

I hereby pledge that I will not make a single EFM edit (feel free to revoke the right if you see me editing them), and if I decide to work on them after gaining some experience, I will ask for an additional permission here. I'd normally go for a temporary access, but the filters are constantly developed, and I anticipate that I'll have to check them regularly in order for us to be in synch.

Cheers

Vito Genovese 16:53, 8 January 2010 (UTC)

The private English Wikipedia filters would be of little utility for any other project. Ruslik_Zero 17:39, 8 January 2010 (UTC)

Perhaps the general structure and makeup of the filters would be instructive. I am minded to grant this request, but will wait for additional opinions. –xeno^talk 17:43, 8 January 2010 (UTC)

Vito, as Xeno said the private filters on this Wiki would not be very useful as they target specific vandals and such. If you need it then I would be glad to help you develop filters on your wiki. Triplestop x3 17:56, 8 January 2010 (UTC)

Ruslik said that, I disagreed =) –xeno^talk 17:57, 8 January 2010 (UTC)

Quite a few of the more "general" filters are private, to my surprise - 72 or 82, for example, which are relatively basic things and don't rely on keywords or language-specific material. Added to that, being able to examine a big set of existing complicated filters - and private ones tend to be more complicated and special-caseish - is the sort of thing that helps immensely when trying to write your own ones. I can see real benefits to letting trusted people use our experience to help build their own systems - there's no reason to force them to re-invent the wheel every time they want to construct something complicated. And I think we can trust him :-) Shimgray | talk | 18:01, 8 January 2010 (UTC)

(ec) A number of private filters might be useful at the very least for an initial setup, like for some random example Special:AbuseFilter/34.
I've briefly talked to Vito before when he ported easyblock and other scripts to tr-wiki, and I too would grant this request. Amalthea 18:02, 8 January 2010 (UTC)

Well, there is no harm with him being able to see the filters, however I'm not quite sure if the same patterns of vandalism prevalent here are also widespread on other sites. Triplestop x3 18:06, 8 January 2010 (UTC)

I can definitely see the value of it, even if the filters are en-specific. There are enough filters here to be good examples and "documentation through code." I agree, we should enable access. tedder (talk) 18:58, 8 January 2010 (UTC)

I also agree that there are no trust issues, and that even if some of our specifics aren't useful, they demonstrate how to do certain things, and some of them may be relevant. Unless someone has a problem with granting them, I will in a little while (or someone else can). It's not really a big deal in my personal opinion. :) Ale_Jrb^talk 19:07, 8 January 2010 (UTC)

As usual, I oppose giving anyone access who doesn't intend to use the rights. How about just giving Vito Genovese the text of the filters... Would that not be easier? Prodego ^talk 19:52, 8 January 2010 (UTC)

What's the tangible problem with giving the rights in this case, Prodego? I see absolutely no reason not to help out a sister project in this way (so support giving him the permission). The editor is clearly a trusted user over there, and has pledged only to copy the filters. Sounds reasonable. ╟─TreasuryTag►Woolsack─╢ 19:53, 8 January 2010 (UTC)

I oppose it because we have declined several other requests to 'copy filters'. You do not need our filters, you can write your own. There are more than enough public filters to get the idea of what we are doing, and enwiki had no filters to 'copy', so clearly there is no need to copy anything to get a working system. Prodego ^talk 19:58, 8 January 2010 (UTC)

But that's just creating needless difficulty. The whole point of Wikipedia is that it is open-source and champions the concept of free content – obviously, the human mind is capable of producing this from scratch because it has done in the past; but there's no point forcing it to when the material already exists. It's against the spirit of free content, and simply unfriendly, to demand a sister project to build up their anti-abuse arsenal from scratch. ╟─TreasuryTag►cabinet─╢ 20:00, 8 January 2010 (UTC)

He will indeed use the right, he simply won't edit the filters. Since we can't grant abusefilter-viewprivate separately, he is requesting EFM. And no, I don't think taking the time to cut and paste all the private filters and emailing them to him is easier than flipping a switch in the userrights screen. –xeno^talk 20:02, 8 January 2010 (UTC)

He does not need private filters, there are more than enough public ones. Private ones are private for a reason, if he is trusted enough on enwiki to see them, then this is the place to go. Prodego ^talk 20:08, 8 January 2010 (UTC)

Where's the harm? –xeno^talk 20:09, 8 January 2010 (UTC)

You know what, just go ahead and do it, there will be no harm from his access. The only problem is if everyone decides they are unable to write filters themselves and everyone finds the public filters so inadequate that they need rights on enwiki that they will never use. I would not expect to be granted them with such a weak request on trwiki, or anywhere else, and would not do the same. Prodego ^talk 20:17, 8 January 2010 (UTC)

He's not just an admin, but a crat on another (magnitude 5) project. I think we can fly the AGF flag a little bit. And it's not like he's asking to copy our homework, there's no harm in letting them learn from our filters. –xeno^talk 20:22, 8 January 2010 (UTC)

That's great that trwiki trusts him as a crat. But it really has absolutely no weight here, because this isn't trwiki, this is enwiki. Prodego ^talk 20:27, 8 January 2010 (UTC)

Which might be relevant if he were at WP:RFA. He's not. I must admit, I'm not quite sure what the actual problem is. Ale_Jrb^talk 21:26, 8 January 2010 (UTC)

"The only problem is if everyone decides they are unable to write filters themselves and everyone finds the public filters so inadequate that they need rights on enwiki that they will never use." - How, exactly, is that a problem? Heaven forbid that smaller projects want to have the same level of abuse prevention that we have. Saying that he shouldn't be allowed because we can't trust him is understandable (silly given the situation, but understandable), but saying that trwiki can't use our private filters, because then other projects would want them is like a little kid not willing to share his toys. If cluttering up the list of users with EFM is that much of a problem, he could just be granted it temporarily. Mr.Z-man 22:38, 8 January 2010 (UTC)

Prodego, I feel I must echo Treasury's statement above. Wikipedia is not only open-source, it is one of the preeminent champions of open-source content. The filters are not private so as to protect our IP, they are private so as to prevent vandals from seeing them and altering their behaviour to suit. The notion that we would deny a sister project access to something that has been beneficial to us runs entirely counter to our core philosophy. There appears to be no reasonable chance of abuse in this instance, and it is our ethical duty to help out our sister projects. My !vote doesn't carry any weight here, but it's firmly in support of Vito. Throwaway85 (talk) 07:53, 10 January 2010 (UTC)

Couldn't agree more. All rights should be easy-come-easy-go, and be granted whenever there is a need and no reason to distrust. EFM has grown to be a flag with a lower threshold than admin, so pointing to RfA is disingenuous and a total strawman. Being a crat on trwiki carries exactly as much weight as being a crat on enwiki: whatever weight you choose to give it. The only acceptable reasons to refuse EFM are a lack of trust that the confidentiality of private filters will be maintained, or a belief that they will not use any of the permissions granted with EFM. The latter is quite manifestly not true, as he will use the editfilter-viewprivate permission. The former is, IMO, quite ludicrous, but is a position you are entitled to take if you wish. Is that your intention? Happy‑melon 13:13, 11 January 2010 (UTC)

I think we've reached a pretty clear consensus here, and I don't see much benefit to dragging this on, so I've gone ahead and granted the request. Ale_Jrb^talk 09:02, 13 January 2010 (UTC)

Thank you guys for the comments. The right will be used extensively, and will be of great benefit to tr.wiki. Again, on behalf of Turkish Wikipedia, thank you.

Vito Genovese 03:13, 17 January 2010 (UTC)

Not a problem, hope the filters help! Throwaway85 (talk) 03:32, 17 January 2010 (UTC)

High condition limit hit rate

Every so often lately it seems there have been swells of hitting the condition limit. Recently I've seen things as bad as

Of the last 3,689 actions, 25 (0.68%) have reached the condition limit of 1,000

However usually it's not quite that bad (usually I see it down near 1 out of every 2000, but that still seems bad to me). I'm not sure if this is a problem or not. I fear this may be, at least partially, due to filter 278, which was recently developed. Then again, this might be the placebo effect at work here, considering I personally know I was working on this as recently as yesterday. I can't make much sense of it, because, even after having not been changed for a day, the consumed conditions bounces around periodically, but then stabilizes around 10-20. I've made a cursory look at other filters and don't see anything that looks to be poor. Does anyone have any insight into this (potentially non-existent) problem? --Shirik (Questions or Comments?) 02:32, 18 January 2010 (UTC)

Since making an optimization to filter 278 this problem appears to have gone away (though it may be premature to make that assertion). However, a second opinion may still be warranted. --Shirik (Questions or Comments?) 05:20, 18 January 2010 (UTC)

0.68% is not a high level (> 2% is high). Ruslik_Zero 05:25, 18 January 2010 (UTC)

Fair enough; it was higher than what I had seen in the past (my limited past), so I was a bit worried I might have been causing a problem. Glad to know it wasn't as bad as I had feared. --Shirik (Questions or Comments?) 05:50, 18 January 2010 (UTC)

2% is the level where (IMO) something has to be done. Typically I do cleanups when it gets close to that. Prodego ^talk 01:42, 19 January 2010 (UTC)

Bring Filter 200 Back online

Previous discussion: Wikipedia talk:Edit filter/Archive 4#Filter 200, or should the EF be engaged to track non-abusive, non-"wrong" edits?

Hey, I was hoping to get Edit filter 200 back online. 200 is the filter that tracks when Prod templates are removed. I would like to take a look at these, and determine weather to bring the articles to AFD. It would be very helpful. Ill drop a line at PROD VPM and the PROD Wikiprojcet. Thanks, Tim1357 (talk) 23:43, 18 January 2010 (UTC)

The original reason it was deleted was because of the general concept that abuse filters were not intended to assist bots. If there's general agreement that this will be used manually then I don't really have any objections against it, but it may go against the concept of edit filters being primarily for abuse (and removal of a PROD template is certainly not abuse). --Shirik (Questions or Comments?) 01:07, 19 January 2010 (UTC)

I still think it's a bad idea and a bot can be written to monitor the RC feed rather than bog down the edit filter and add entries to users' filter log for perfectly legitimate edits. –xeno^talk 01:43, 19 January 2010 (UTC)

Wouldn't it be easier to have a bot watch what is in the prod category and note when it changes? Prodego ^talk 02:24, 19 January 2010 (UTC)

I agree I think use of a bot is both easier to detect and better for performance. --Shirik (Questions or Comments?) 02:40, 19 January 2010 (UTC)

Sure, that works too. –xeno^talk 17:11, 19 January 2010 (UTC)

• I don't think I really care one way or the other, but this strikes me as kind of pointless. If a user adds a PROD, but doesn't bother to watchlist the article and keep up to date on it, then they didn't really care that much anyway. Anytime anybody removes a PROD that I have added without fixing the problem or explaining their reasoning, I go straight to AFD, as I'm sure most other users would in that circumstance. Beeblebrox (talk) 05:06, 19 January 2010 (UTC)

I was under the impression that an edit filter was more efficient. Isn't there, too, a way to make it so it does not advertise "Tag: prod contested or removed", rather populate a Recent Changes list? In response to Beblebrox, the list would be helpful for those who do not watchlist their prod, so others can monitor it for them. Tim1357 (talk) 11:50, 19 January 2010 (UTC)

Depends on your definition of 'efficient'. To assess whether something has changed, a bot just has to grab the last two revisions of every page that changes. That might sound like a lot, but when you consider the thousands and thousands of read requests we can get per second, it isn't really a big deal. The abuse filters, however, have quite a strict condition limit, so it's better to save that for filters that can't be done so easily with a bot. Ale_Jrb^talk 13:43, 19 January 2010 (UTC)

I agree that this should be done by a bot, if at all. IMO, given that we have a condition limit, tag/log-only filters should be kept to a minimum. As Prodego notes, a bot could do this without having to look at the article at all by checking the prod categories periodically. Mr.Z-man 17:08, 19 January 2010 (UTC)

• I should point out that another reason for wanting to see deprodded articles is to improve them, not just delete them. Or are only deletionists interested in this? Fences&Windows 20:15, 19 January 2010 (UTC)
• Myself, when I add a prod, unless there is some special reason I think the article must be removed, I accept the fact that it might not be; when I started here I kept a list, but I stopped long ago because of the inability to follow up. I also patrol prod, and remove about 5 prods from each daily batch, often by changing to redirects, sometimes by speedying them, and I see that another 10 or 15 are removed or at the end declined, by other patrollers, most of whom know what they are doing. I'd accept a filter listing for where a prod was removed by an ip or the creator--those are the ones where deprodding can be a problem. DGG ( talk ) 01:50, 20 January 2010 (UTC)
  • That version of the filter seems like a good middle ground; when it's the author removing the prod, it can indicate other problems such as SPAs or spam. It can also indicate where new editors could use some assistance with our policies. In either case, highlighting the issue is of value. UltraExactZZ ^Said ~ Did 15:53, 20 January 2010 (UTC)
    • It is not possible to identify the creator of the article once it has been created with the abusefilter. Prodego ^talk 15:54, 20 January 2010 (UTC)