On Thu, 5 Sep 2002, Ricardo Kustner wrote:
> So my guess is that it could be that the process of cleaning a whole
> site with hundreds of pages from the cache (using
> http://www.cache.dfn.de/DFN-Cache/Development/Purge/, which finds the
> objects in the file cache and constructs the PURGE requests)
> sends a lot of PURGE requests in a short time; meanwhile the websites
> are still being accessed by clients so maybe something goes wrong,
> causing some of the PURGE request to get lost?
Maybe.. as I said, see your access.log.
> I'm just guessing here... it's kinda weird and hard to reproduce.
> It turns out that a Mozilla-reload *always* makes squid serve a fresh
> document (no need for a PURGE) so that seems to be unrelated.
Mozilla sends a "Cache-Control: max-age=0" on reload, which forces Squid
to bypass the cache.
Most other browsers send a "Pragma: no-cache", which can be overridden by
refresh_pattern by the "ignore-reload" option, and I seem to recall you
are using this option.
Some (most?) versions of IE does not send a "Pragma: no-cache" or any
other "please do not give me a cached copy" header unless configured to
use a proxy... which btw is a valid thing to do, if it wasn't for
surrogate servers/accelerators or the black art of "transparent
interception" proxying that messes up the assumed
user-agent<->originserver relations.
Regards
Henrik
Received on Thu Sep 05 2002 - 14:46:45 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:10:07 MST