Henning Brauer writes:
> On Thu, Dec 20, 2001 at 04:04:50PM +0000, Dan Melomedman wrote:
> > Henning Brauer writes:
> > >> * big-todo
> > > gains nothing. That's a patch for a very rare very special circumstance.
> > Could you explain? I am not familiar with the patch, but it should be there
> > for a reason.
>
> It's for the rare situation where you nearly always have a high number of
> not-yet-preprocessed messages ("high" does not mean a few hundred here!). In
> normal cisrcumstances it doesn't help at all. More possible slows down
> things a bit.
Yes, but how rare are those circumstances? After all, Russell Nelson
must have had a reason to write the patch...
Could it happen under normal usage? DoS? OS bug (I don't recall
exactly, but I think I saw it happen w/ Solaris ;-)? Too high a
tcpserver concurrency?
And once you get there, isn't manual operation (slow the injection
rate a bit) required to bring things back to normal?
As I'll be using a filesystem that does linear directory reads, I'm
worried that thousands of messages under todo/ could make qmail
choke.
Anyway, I just want to play it safe, and it is much easier to apply the
patch before the server goes into production.
Sorry to ask again, but do you really think one can do without this
safety net?
Thanks,
--
Adriano
