Problems with the wiki / FAQs.

March 29, 2005 2:26 PM

The Wiki/FAQ has been pornified. I'd revert it to a previous version but I can't seem to figure out how. Could someone who knows how this stuff works please revert it and then let us know how it's done? (AskMe or Meta, when in doubt take it to Meta)
posted by Mitheral to MetaFilter-Related at 2:26 PM (27 comments total)

I just did, before you posted this, oddly. You don't have to log in to do it, that's why it's so easy to pornify. :)
posted by Ethereal Bligh at 2:35 PM on March 29, 2005

Oh, if you're asking specifically how to "revert"...there doesn't seem to be that functionality (based upon my very cursory search). I just cut-and-pasted the most recent non-spammed version.
posted by Ethereal Bligh at 2:37 PM on March 29, 2005

But what did you click to get it to revert? I managed to find the previous versions but I couldn't make that version the current one.

On preview, Ah! I was looking for some link to follow.
posted by Mitheral at 2:39 PM on March 29, 2005

Wow, the Viewropa entry had been messed with 900+ times.
I just reverted it back to "Save 1" -- I fear some more recent updates have been killed, but it's preferable to what replaced them.

Perhaps some sort of registration should be put in place?
posted by me3dia at 2:51 PM on March 29, 2005

If only there was a "Pornify" button on my keyboard or on my remote control.
posted by Stan Chin at 2:52 PM on March 29, 2005

You don't have one of those, Stan? I do—I couldn't live without them.
posted by Ethereal Bligh at 2:57 PM on March 29, 2005

I think it's one of the buttons above the numeric keypad.
posted by Bugbread at 3:00 PM on March 29, 2005

I just went and banned the source of the attacks (I think). In any case it wouldn't have been long before RichardP's magical spam-killing bot went and fixed everything and banned the bad guys. People who spam wikis deserve a special place in hell...
posted by adrianhon at 4:07 PM on March 29, 2005

see instructions here for how to get rid of this.

does richardp use a bot?! i thought he was just obsessive! :o)
posted by andrew cooke at 4:22 PM on March 29, 2005

Yeah, it's definitely a bot. I remember thinking for a while, 'holy shit, this guy likes to maintain wikis.' Then I thought about it some more :)
posted by adrianhon at 4:42 PM on March 29, 2005

when the site was down on Sun, i noticed it too--weird lists of porn sites.
posted by amberglow at 4:45 PM on March 29, 2005

guess so. i sometimes fix a few to help out. won't bother now!
posted by andrew cooke at 4:49 PM on March 29, 2005

If only there was a "Pornify" button on my keyboard or on my remote control.

The next best thing. (NSFW!)
posted by Wet Spot at 5:47 PM on March 29, 2005

Hey, has anyone thought of using photos of celebrities (have the human identify them) as an anti-spam scheme? 'Cause I just did. I first thought about pulling Google images (and looking for dupes and hoping it really was a photo of the celebrity you searched on), but then thought that using IMDB photos of celebrities would be better. And do a combination, like three or four. And just ask for the first initial of their first names.
posted by Ethereal Bligh at 5:54 PM on March 29, 2005

Not great for international users, necessarily (unless it comes with an instruction sheet).
posted by Tlogmer at 6:21 PM on March 29, 2005

That would suck for me, because most of the celebrities I know are all mid-90's and earlier (I know the names of the folks from afterwards, but I couldn't tell Ashton Kuscher from any of N'Sync)
posted by Bugbread at 6:22 PM on March 29, 2005

If only there was a "Pornify" button on my keyboard or on my remote control.

You know we have a Sleazy Button now, don't you Stan?
posted by Smart Dalek at 6:47 PM on March 29, 2005

Well, if you didn't recognize it, you could resubmit and get a different celeb. If they were chosen to be the most famous, then most people would quickly get someone they recognized.
posted by Ethereal Bligh at 7:32 PM on March 29, 2005

With multiple choices, that makes sense. I think we could all identify George Bush or Michael Jackson, regardless of our respective nationalities.
posted by Bugbread at 8:00 PM on March 29, 2005

I think that people who don't watch movies often or TV at all (like me) would be able to identify maybe 10% of celebrities, at best.

On preview - you'd need a fairly large number number of images (greater than 2) in order to keep non-humans out. Notice that most sites which use distorted letters generate them dynamically - you have to rule out any possibility of a brute-force attack, and force the attacker to rely entirely on skills present only in humans at the current stage of AI (in these cases, reading, or facial recognition).
posted by advil at 9:16 PM on March 29, 2005

you could generate celebrities automatically too - like that ashlee simpson woman.
in fact, if most celebrities were being generated by computers to provide access contol to web sites it might explain a lot.
posted by andrew cooke at 2:48 AM on March 30, 2005

Case in point. I know who Ashlee Simpson is (from the MeFi posts about lipsynching and getting booed at a halftime show), but I don't know what she looks like.

Perhaps a better choice would be images? A toaster, a bicycle, a clown, and to make it extra metaFiltery, an iPod or a Moleskine (no, not seriously).

And probably making it multiple choice would be better, so you don't have people frustrated by seeing a clear picture of a bike but not being allowed access if they type "bike". I don't know how much easier that would make botting, though.
posted by Bugbread at 3:57 AM on March 30, 2005

or, for the devout among us we could use images of saints. or some kind of multiple answer scheme - some would answer "pincushion", some "saint sebastian", for example.
posted by andrew cooke at 5:43 AM on March 30, 2005

Way to completely shut the door on our blind brothers and sisters, people.

Er. Do we have any?
posted by beth at 10:48 AM on March 30, 2005

No problem; an alternate login system would play a noisy sound recording of an easily recognizable animal.
posted by nobody at 2:11 PM on March 30, 2005

Ooh, that'd be good! I do think that, all things considered, very common objects (and animal sounds) are better than celebrities. The advantage of celebs is that there's no ambiguity about their names. There is often ambiguity with object names. So you'd need to be very careful what you pick and yet still pick enough so that the system is useful. But that's possible, I think. You lockout IPs that repeatedly hit the system in a way that a person who got one wrong and tries again (and maybe a third time) wouldn't. So I think image or sounds as an alternate wold work well. Note that, per beth's objection, the current systems that ask for an image of an obscured number are also non-sight-disabled friendly. The inclusion of nobody's sound alternative would be more inclusive than average! Also, the number identification schemes are getting close to becoming not very useful as spammers are starting to use much more sophisticated technology to identify numbers in an image. I think it'd be a good long while before you'd be able to reliably identify common objects. That's a big, continuing problem in computer vision after all. Numbers obscured somehow is a very much more limited problem domain. But identification of very common objects or sounds would be slightly more difficult, but still very easy, than identifying these obscured numbers have been.

The problem requires something that can't efficiently be automated. Another approach I thought of was to have a little app that people downloaded that used crypt tech to produce a valid number from an input number (that your website would provide). The app would be freely available to the website's users, but the answer/response would be specific to a specific website's app. That could be automated, of course, by a spammer that wanted to specifically spam your website; but it would raise the bar for their effort expended high enough to not make it worth their while to do so (unless you're a very, very high-profile site).
posted by Ethereal Bligh at 2:49 PM on March 30, 2005

"But identification of very common objects or sounds would be slightly more difficult, but still very easy, than identifying these obscured numbers have been."

...for people, but much harder to automate. Was my point.
posted by Ethereal Bligh at 2:51 PM on March 30, 2005