Chad Perrin: SOB

2 March 2007

Spam Catch of the Day: Catch and Release

Filed under: Geek,Metalog — apotheon @ 05:46

It’s imperative that, when you design a spam filter, you give it an effective interface for sifting through everything caught in the filter to be sure there aren’t any false positives.

Unfortunately, for WordPress there seem to be exactly two options for spam filters: the built-in moderation interface and the Akismet spam filter plug-in. The former has an adequate (but not excellent) interface, but it entirely incapable of filtering trackbacks (now that is a bone-headed design decision). The latter filters everything, but its interface is complete crap. I rather suspect that I’ve missed a few false positives over the last couple months, since I started using Akismet again because of a rise in frequency of trackback spam. Every once in a while, a spam comment slips through to the moderation queue — and every once in a while, something that isn’t spam is “caught”.

Unlike the Akismet plugin’s interface, the moderation queue allows me to identify someone that posts as a non-spammer, after which that person’s comments will never appear in the queue again. Unlike the Akismet plugin’s interface, I can delete one page’s worth of spam comments in the moderation queue — the Akismet plugin only allows me to delete the entire contents of the spam filter in one shot (or nothing at all). In contrast with the Akismet plugin, the moderation queue allows me to “manually” add something to the moderation queue, and delete or de-spam posts without affecting whether they are used to increase the accuracy of the queue if I so desire.

It’s easier to simply skim a page quickly and recognize at a glance whether there are any non-spam comments in the moderation queue than in the spam filter for the Akismet plugin, too.

Today’s spam catch is a false positive — a catch and release. It’s the first comment on the “In the news: do NOT fuck with American tourists” post.

All original content Copyright Chad Perrin: Distributed under the terms of the Open Works License