SpamAssassin development is still active

Posted Sep 1, 2017 19:49 UTC (Fri) by dskoll (subscriber, #1630)
Parent article: Spam filtering with Rspamd

SpamAssassin may not have had a release in a while, but its SVN repository is still active, with the most recent commits less than a day ago. Most of SpamAssassin's development happens around the rule sets, which are updated outside the normal release cycle.

(Log in to post comments)

SpamAssassin development is still active

Posted Sep 8, 2017 8:09 UTC (Fri) by madhatter (subscriber, #4665) [Link]

Thanks, that was the point I thought to make, also. You having kindly made it, I went back and looked at the output from my nightly sa-update job to see how often the rulesets were getting updated; there have been 52 updates so far in 2017, the most recent of which was on 24/6.

They seem to come in waves - no updates for several weeks, then updates every night (or nearly so) for about a couple of weeks; there have been four such waves this year so far. I'm guessing it's part of an incremental development model where they wait until there's good reason to release new rulesets, then iterate towards a maximally-effective version over several nights, but that's just a guess.

At any rate, I think you're very right to make the point that SA's decoupling of the rule-running framework from the rules themselves means that looking only to updates of the framework to determine whether the project is still evolving will give an inaccurate answer. 52 releases this calendar year doesn't look moribund to me.

SpamAssassin development is still active

Posted Sep 8, 2017 21:13 UTC (Fri) by nix (subscriber, #2304) [Link]

They seem to come in waves - no updates for several weeks, then updates every night (or nearly so) for about a couple of weeks; there have been four such waves this year so far. I'm guessing it's part of an incremental development model where they wait until there's good reason to release new rulesets, then iterate towards a maximally-effective version over several nights, but that's just a guess.

It's actually part of a model where Apache-side changes kept breaking the score generation, and it was mostly undocumented so people had to figure out how it worked again and fix it. Also, there are only just enough emails in the corpus, so it keeps on dipping below the threshold and throttling. More volunteers needed!