"Jason R. Mastaler" <[EMAIL PROTECTED]> writes:

> Tim Legant <[EMAIL PROTECTED]> writes:
> 
> > I'd go with something around 10, but see below.  If a real person
> > sends me 10 email messages in a row without confirming one, sending
> > 40 more confirmations (assuming s/he keeps sending mail) probably
> > won't make a confirmation response any more likely.
> 
> This rests on the assumption that the TMDA user is using
> CONFIRM_APPEND though, correct?

I didn't think so, but maybe I'm missing a connection here...?

> If they aren't using this, I could imagine cases where someone would
> send 10 messages expecting to get back 10 confirmation requests.

Without responding to any of them?

> Also, this limit is for _all_ auto-responses (confirm requests,
> confirm accepts, and bounces all add to the total).  So, a limit of 10
> would be reached after a sender confirmed only 5 messages (5
> confirmation requests + 5 confirmation acceptance notices returned).

I'm curious why you want to do it this way.  I don't see the point,
right off, of counting acceptance notices.  Do we gain something
versus counting just confirmation requests?

> Also consider shared setups where multiple recipients share the same
> TMDA configuration and queues (such as under a qmail virtualdomain,
> vpopmail, etc).  This would raise the necessary limit slightly.

Yeah, this is a different situation.  I'll have to think about this
one.

> > If, upon receipt of a single confirmation message, all messages for
> > that sender are released, the need for a high threshold is reduced,
> > it seems to me.  The later messages, for which a confirmation wasn't
> > sent, will still be successfully delivered.
> 
> This sender-based pending queue is an independent idea, and one which
> will require quite a bit more time to design and implement.  The queue
> would have to be carefully redesigned, and related tools like
> tmda-pending would have to be trained to accommodate.

If this requires the sender-based queue, I withdraw the suggestion.  I
wasn't thinking of that full implementation, though.  Perhaps that
*would* be necessary to implement releasing all messages from a given
sender.  Scanning all messages in pending/ would also work, but might
be prohibitively expensive... hmmm.


Tim
_________________________________________________
tmda-workers mailing list ([EMAIL PROTECTED])
http://tmda.net/lists/listinfo/tmda-workers

Reply via email to