[ZODB-Dev] ZODB for spambayes server-side filter?

[Tim Peters]

[Simone Piunno]
> I'm working on a server-side spam filter based on spambayes.
> After some prototyping with BDB4, I've started to look at ZODB.
> I'm trying to understand if this is a good idea.

FYI, I started the spambayes project, and the initial design was made with
ZODB's BTrees in mind for the primary mapping structure.  The code Jeremy
referenced lives in the spambayes project's pspam directory.  It's probably
suffered significant bitrot, though (nobody has had time to look at it for
months and months).

However you collect training material, I suggest that you funnel all
training into one thread per training database, and do training on batches
of messages rather than one at a time, committing once at the end of a
training batch.  That will cut down on the number of conflicts you're faced
with.  There will be contention between training and scoring, since, for
example, virtually all messages contain words like "the", and synthesized
features like "has a Date header".