> On Wed, 8 Jan 2003, James P. Roberts wrote:
>
> > is there any simple way to do something in the code, to achieve the
> > following:
> >
> > (a) if we get some number of timeouts in a row, to a particular
DNSBL,
> > temporarily stop using that DNSBL.
>
> No. Difficult to define "in a row". This is because Exim is not a
> centrally controlled application; independent processes are not
> coordinated. (I'm giving a talk about this in Sweden next month,
> pointing out what you gain and lose by such "anarchy".)
>
> To do what you want would require Exim to write a hints record for
every
> timeout, and to read a hints record before attempting the next DNSBL.
I
> guess this wouldn't impact too much, because it would be mostly
reading,
> which does not require exclusive access to the file.
>
> WishListed.
>
Understood. I was just simple-mindedly hoping for a way to do what I
would otherwise have to do manually (and therefore less
effectively/timely). Feel free to delete this from the wishlist, at
least as far as I am concerned.
> > Another idea balloon to float:
> >
> > If using +defer_unknowns with one's DNSBL list, and if only one of
> > multiple DNSBL is giving timeouts, and the other(s) are giving a
clean
> > bill of health, do not impose the defer.
>
> WishListed.
>
> <smallprint>
> "WishListed" means I've put the idea on the list for thinking about in
> the future. What gets implemented (if anything) may not be exactly
what
> is suggested.
> </smallprint>
>
> Philip
>
> --
> Philip Hazel University of Cambridge Computing Service,
> ph10@??? Cambridge, England. Phone: +44 1223 334714.
>
Thanks, Philip. Trust me, I understand the "Wishlist" concept! I
appreciate your consideration of the ideas.
If I had a preference, it would be for the second idea above, regarding
the change in behavior of "+defer_unknowns." After the recent experience
with a DNSBL outage, I don't feel comfortable turning that option back
on, as is.
The first idea is just a wild idea intended to stimulate thought, and
perhaps lead to a much better solution. Hopefully, the second idea
presented is such a thing. I remain open to ideas. I like to get
people to put forth a whole bunch of ideas, most of them "crazy" in the
context of the problem to be solved, in hopes of one "crazy" idea
leading someone to think of a truly good solution. I love nothing
better than a good brainstorming session! (OK, wait, I take that
back... there are a couple of things in life that I like better! But
those would be "off-topic." *grins*)
Jim Roberts
Punster Productions, Inc.
p.s. - Just a quick thank you for the very clear log messages, which led
me to pin-point the problem within a few minutes of realizing there was
a problem. The fix was then a simple change to one line in exim.conf.
Well-written software, if there ever was any. Thanks!
