Re: Anyone working on mail/crm114 port?

[Robert McMeekin <rrm3_(_at_)_rrm3_(_dot_)_org>]

* Andreas Kahari <ak+openbsd_(_at_)_freeshell_(_dot_)_org> [05/09/04 00:15]:
> On Sat, May 08, 2004 at 06:49:54PM -0400, Robert McMeekin wrote:
> > On Sat, 2004-05-08 at 23:20 +0100, Andreas Kahari wrote:
> > > Is anyone working on a port of the spam crm114 spam filter (and
> > > the "TRE approximate regex engine" on which it depends)?
> > 
> > After building it a month or so ago, I couldn't see why anyone
> > would want to use this.  It won't run on a 166MHz/32MB machine
> > without the addition the minimum '-w' option to the #! line in each
> > rule file, and at that it still takes hours for it to process a
> > simple rule, let alone those recommended for it's mail filtering
> > system.  What have your results been like, and what kind of machine
> > are you running it on? 
> 
> I've been using cmr114 as a personal mail filter with procmail
> (i.e. not system wide) since February on two OpenBSD/i386
> machines (both around 2GHz/512Mb) and on one multi-user
> NetBSD/alpha (466MHz/512Mb) system.
> 
> I know it eats memory, which probably makes it a pain to run on
> the types of machines that you mentioned, but it's very much
> faster (and more accurate) than e.g. SpamAssassin.

Well, after trying, unsuccessfully, to run the tests that come with the
crm114 tarball, I decided not to even try to filter mail with it.
Anyway, I think a port should include some kind of warning about
memory requirements.  I mean, IIRC the thing defaults to something
like 4 x 24MB filter files in memory, and the minimum is something
like 4 x 8MB or something like that.

> The NetBSD/alpha system (www.freeshell.org) logs procmail usage,
> and the average procmail processing time per message, for all my
> incoming e-mails since the beginning this month (1329 of them),
> is wallclock=2.9s user=0.5s system=0.1s.  On the i386 machines,
> this is faster.

Hey, that's not so bad.  I don't how well it would work with every
user filtering their mail with it, but those numbers certainly put it
as a decent personal mail fiter.

Also, wasn't there something with wchar.h so libtre doesn't support
unicode and thus crm114 doesn't filter unicode messages, or something?

-- 
Robert McMeekin <rrm3_(_at_)_rrm3_(_dot_)_org>