Re: Fwd: [ANNOUNCE] : Filters patch against 1.2.0 [e.allaud@wanadoo.fr]

[Brian Stafford <brian stafford uklinux net>]

On Wed, 26 September 00:05 Carlos Morgado wrote:
> On Wed, Sep 26, 2001 at 12:12:59AM +0200, Pawel Salek wrote:

> > The issues that I think should be discussed:
> > 1. should balsa filters be stored, or have an export option to sieve 
> > syntax, to allow easy upload to remote mail servers?
> > 
> if possible yes. if possible write libsieve (as everyone seems to agree on
> sieve) and make it Good.

I think the CMU sieve might already be libsieve - I'll check that when I
get a moment.

In any case there should be libsieve with a decent API and a deliver/procmail
style program which uses it.  That would provide for apps which need tight
integration with the filtering (e.g. balsa) and for those where the integration
is looser, i.e. sendmail etc. delivery processing.

> > 2. how can we stimulate usage of sieve server-side mail filtering? I think 
> > availability of server side filtering is crucial. I use it now presently 
> > to keep my primary inbox free from discussion list messages etc. I know I 
> > could use balsa soon-to-be-available filtering to hide bulk mail from the 
> > inbox but the fact is, I cannot always use balsa to access my mail for 
> > various reasons.
> > 
> 
> the imap argument is moot imho as the imap server itself should do the
> filtering (enter sieve) instead of having a mua go about moving messages
> about.

I believe that current Cyrus IMAP has sieve already (better check this too).

> of course, then there is the weird cases that actually need to sort imap for
> some reason or other and then the MUA is king.

This is why I feel that filtering should be provided in a UA regardless of
its availability elsewhere in a system.

> now, speaking *for* procmail it managed to implement server side filtering
> by becoming a MDA most MTAs like. 

No reason the scripts should not be sieve instead.

> anyone volunteers ;)

I got as far as writing a sieve parser in yacc about 18 months ago.  Sadly
that project fell on stony ground (in part because I was working from internet
drafts and they were in a state of flux at the time).  However, assuming the
licence is OK, there might still be the CMU sieve implementation.

Brian