On 07/08/12 23:31, Todd Lyons wrote:
> So I saw on a Freecode daily email that a new release of rspamd was
> available. It claims to have 10x better performance than spamassassin
> with the same rules. Has anybody ever used it?
>
> This patch is in their source to get it working with exim:
> https://bitbucket.org/vstakhov/rspamd/src/649c61b89b43/contrib/exim/patch-exim-src_spam.c.diff
>
> It looks like good code to me, so with a little bit of #ifdef we could
> make rspam vs spamassassin a compile time setting. Or maybe add a new
> "rspam" condition that calls this code instead of the existing "spam"
> condition code.
>
> Comments?
>
> ...Todd
Never used it, but it certainly looks interesting. I've added it to my
"test asap" list, but that's a loooong list ;)
The spamc protocol has been expanded to include a number of other
indicators that can be passed to the rspamd daemon like the ip and helo,
but for the most part looks to be a straight slot-in replacement.
Except for those added bits, and only at a casual glance, the existing
spamassassin would need very little modification as shown by their patch.
Their provided patch doesn't allow for side by side use as it replaces
the existing spamc implementation, including keyword. It also add a new
function "push_line" which doesn't seem to fit with how any of the other
strings or communications are handled in exim. I'd be more comfortable
with it using the current exim functions, but then again, perhaps their
method is more up to date and the rest of exim should be updated too.
The claim that it is 10 times faster than SpamAssassin would be an
awesome start. If it was equally memory efficient it would be even
better. I did find some concern voiced about whether the rspamd was able
to be using uncredited SA rules and be licensed differently than SA.
Those were from 2009 though and I haven't looked to see if anything has
changed.
Ted.
