"Michael T. Babcock" wrote:
>
> ----- Original Message -----
> From: "Dave Sill" <[EMAIL PROTECTED]>
>
> > The problem you seem to have with qmail-send is that it processes
> > deliveries in the order in which they're received. This property is
> > known as "fairness", and it's perfectly reasonable in a queuing system
> > that doesn't support multiple priorities.
>
> Actually, its just FIFO. Fairness would involve queuing and re-ordering of
> some form. Fairness infers intelligence on the part of the queuer. Now you
> can say that there is fairness built in to the randomness of the queue's
> order, but its not "fairness" in a queuing sense. SFQ comes to mind w.r.t.
> IP packets.
>
> Incidentally, I don't agree with the original poster's assumptions. But
> qmail-send isn't "fair" -- thats a level of complexity that wasn't added and
> won't (probably) be added because of possible implications for security and
> stability.
Stability is something I could see as a major problem. Security could
be a problem if not built in. I think the biggest problems is how
everything will work together while addressing these issues and not
being to complex as to degrade qmail simplicity, which we all love.
I thought about it a little last night and I have been applying what has
been said today. There is allot of issues to consider, however I do
believe something should done to change the way qmail handles queuing to
increase speed. I need to get more information on how things are really
handle, look at some code, and build a plan. Hopefully I will post a
plan next week or even next month for review. In the meantime, I will
just run fourteen qmail processes to get around the queuing problems.
Steven Rice
