Re: Problems with peer misses after 2.1P2 to 2.2S3 upgrade.

David Malone wrote:

> > a) Perhaps you have enabled icp_hit_stale?
>
> I have it commented out, and I think the default is off.

It is.

> > b) Or your peers are using stricter refresh rules than you do.
>
> This is very possible. Has there been some change in the way this is
> delt with between 2.1P2 and 2.2S3?

Looking at the difference betweeb 2.1P2 and 2.2X I notice that there was
a important bug fixed in 2.2.: A 2.1.X remote peer did not obey to the
staleness limits set in the request (using Cache-Control: max-age=X), so
yes it has been changed. Squid 2.2 is much more sensitive to differences
in refresh patterns than Squid 2.1 was. This restores the behaviour of
earlier (1.x) Squid releases. The 2.1 and 2.0 behaviour is incorrect.
Things got broken (different parts of the source out of sync with each
other) some time during the late 1.2.beta developmend phase.

--
Henrik Nordstrom
Spare time Squid hacker