On Wed, 2006-01-25 at 21:43 +0100, Paul J Stevens wrote:

> I don't know sieve... What's the api used there... would that be
> 
> http://www.ietf.org/internet-drafts/draft-ietf-sieve-vacation-05.txt
> 
> ?? I'm reading ... it goes way beyond the vacation(1) solution;
> featurism at large.
> 
> I guess my point is that as important vacation may be, it's not top priority.

It's high up for me ;-)  The only thing I need to make vacation work is
a table to store (message hash, last reply) tuples. I didn't see the
replycache table. I'll take a look at that.

Ok, it's got:

  to_addr varchar(100) NOT NULL default '',
  from_addr varchar(100) NOT NULL default '',
  lastseen datetime NOT NULL default '0000-00-00 00:00:00',

That's good enough! The "message hash" is whatever will distinctively
identify a recipient so that they don't get hammered with vacation
replies. The to_addr works nicely.

I see a lot of code in pipe.c, send_reply(), for getting the addresses
parsed out of the message and then sending the reply. I'd like to
centralize some of that code so that send_reply, send_notification,
send_vacation and perhaps send_bounce (if we reinstate some bounce code)
could be mostly just wrappers around one chunk of code.

> > That would be a nice feature for auto_reply, and the two could
> > reasonably share a table.
> 
> auto_reply does implement rate-limiting, but the minimal reply interval is
> hard-coded (db.c, +4290) and tracked in the dbmail_replycache table.
> 
> Additional parameters such as 'days' can easily be stored per auto_reply row, 
> to
> allow overriding it per auto_reply.

The days parameter comes from the Sieve script.

> And we could expand the replycache table to store all replies sent, not just 
> the
> last.

Not needed.

Aaron

Reply via email to