Re: [squid-users] refresh.c/refreshIsCachable buglet?

From: Andreas J. Koenig <andreas.koenig.gmwojprw@dont-contact.us>
Date: Mon, 03 Jul 2006 10:31:05 +0200

>>>>> On Fri, 30 Jun 2006 17:46:14 +0200, Henrik Nordstrom <henrik@henriknordstrom.net> said:

> fre 2006-06-30 klockan 17:24 +0200 skrev Andreas J. Koenig:
>> I'd be very happy, if the squid team could reconceive this as a buggy
>> behaviour. Isn't it wonderful to improve software by just removing a
>> few lines?

> It's a tradeoff. As the comment says there is not really any benefit in
> refreshing 0-sized objects as the overhead is the same as a full
> request.

> I would think what you want to change is actually the 60 seconds limit a
> few lines earlier... this limit is suitable for normal forward proxies
> in most setups, but reverse proxies may have quite different
> requirements.

Yes, you're right, this solves the problem for *us* and it's nice that
Squid 3.0 has it configurable with the minimum_expiry_time parameter.

I still don't get it why the code lets downstream Squids with default
minimum_expiry_time cache our 1-byte answers but not our 0-byte
answers. This behaviour adds latency and traffic because our not
changing headers are sent over again and again and they are always
over 400 byte.

-- 
andreas
Received on Mon Jul 03 2006 - 02:32:31 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Aug 01 2006 - 12:00:01 MDT