A trio of developments

The latest release of Gallery (2.3) now has Akismet support built in. Thanks to Chris for letting us know.

Kevin at BIGACE CMS has developed a commenting extension which uses Akismet. 

And lastly but by no means least the latest version of vBulletin blog also has Akismet support by default.

Akismet 2.2.1 — Now With Stats!

If you upgraded to Akismet 2.2, you likely noticed a cool new feature: Akismet stats. We’ve now released Akismet 2.2.1, which includes a few bug fixes.

Akismet stats track your comment metrics, including missed spam, false positives (Akismet thought it was spam, but you disagreed), and Yummy Pie, which displays your ham vs. spam count in a nifty pie chart.

What’s ham? The good stuff. Delicious, legitimate comments. Spam is the non-legitimate filler that Akismet excels at blocking.

We’ve also made significant backend changes to speed up Akismet’s response time. If you think a comment is spam (or ham), the system will now adapt instantly to your feedback — a major improvement over the old adaptation time, which could take a few days.

The Guardian on Akismet

The Guardian’s Michael Pollitt says some kind words about Akismet in today’s Technology section, as he notes that Comment spammers never take a holiday.

Mullenweg publishes a graph on Akismet.com which shows comments being exceeded by daily spam peaks of more than 25m. Checking out the spam blocked counter on my blog is satisfying, showing that the spammers are going nowhere fast.

As an experiment, I turned off my spam protection to watch the spam-a-minute flow.

Read the full article here.

Using HTTP 1.1

If you’re using HTTP 1.1 in your Akismet client, or the one in PEAR, there’s a bug that could cause your client to timeout.

The problem does not lie with Akismet servers, the problem lies in an incorrect implementation of HTTP/1.1 persistent connections in the Akismet class. The Akismet API servers support persistent connections for HTTP/1.1 but the class is not written to use persistent connections. Specifying a ‘Connection: close’ header will fix the problem for HTTP/1.1.

Read more about the problem here.