[Mimedefang] BIG problems with mimedefang

[David F. Skoll]

Matthew.van.Eerde at hbinc.com wrote:

> Once you get beyond a certain traffic level the BerkeleyDB
> implementation of the Bayes database becomes a bottleneck and you
> need to move to a SQL-daemon database.

If that's the case, SpamAssassin has a seriously broken BerkeleyDB Bayes
implementation.

In CanIt, we use our own Bayes implementation, and you can choose a
PostgreSQL or Berkeley DB back-end.  We find the Berkeley DB back-end
considerably outperforms the SQL back-end.  However, our
implementation requires *no* locking whatsoever, so that may be why
our Berkeley numbers are better than our SQL numbers.

Even SpamAssassin shouldn't require locking unless the Berkeley DB
file is being written to.  If it's only being read, then concurrent
access shouldn't be a problem.

Regards,

David.