> -----Original Message-----
> > Good update. I did read it several times, but that section was not
> > "clicking" for me before.
>
> Great. So there is no comments on what to improve now?
I think it's great. Anything I came up with for that section would be
trivial nitpicking. If I'm going to pick nits, I'll do it in the form
of a patch.
Dreaming of the future, it would be nice to have section 5, because it
would put people in the right frame of mind to be thinking about this
stuff before they get to it. The rationale behind the extensive use of
callbacks and a few examples would be a great way to bring people up to
speed.
I'm catching up on the "how" but the "why" remains kind of a mystery to
me. But it sounds like the sort of topic I could search the archive for
and get more than I ever wanted.
Also, for better or worse, I've attached the rproxy-based patch to turn
#define READ_AHEAD_GAP into squid.conf-based read_ahead_gap. As
provided, this patch shouldn't change functionality at all unless
read_ahead_gap is set to a non-default setting.
As previously stated, I can't speak intelligently about the value of
changing the read_ahead_gap. I know only that it helps my situation and
may be of use in managing memory for others. I don't intend this in
lieu of finding the actual problem that causes hangs with certain values
in my environment, only as a band-aid until I know enough to do that.
The patch seems stable; I've checked it with cachemgr.cgi to ensure it's
doing what it's supposed to do, and have run approximately 3,000,000
hits through it on 3 architectures with no visible stability problems.
Lucky for me, READ_AHEAD_GAP was only used in the code one time, so this
was an easy patch to verify.
If you want it, you're welcome to it. If you don't want it, it'll live
happily ever after in my local repository and never bother you again. :)
Thanks,
Jeff--
NearlyFreeSpeech.NET
$1/1GB Web Hosting, no minimums, no monthly fees, no kidding.
http://www.nearlyfreespeech.net/
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:15:41 MST