> But wouldnt that only override max-age which is received in headers sent > by servers? > The ones we want to override are from client requests only. > Plus refresh_pattern can not take an acl since it's global and only based > on path.(ie no acls) > > Or am I not seeing things clearly? > refresh_pattern gets involved when updating something already in cache. AFTER the IMS has been actioned. header_access gets involved when sending the request to the server or the response to the client. Thus it was working too late for your earlier config. The closest thing I know of squid offering is the 'reload_into_ims' or 'refresh_stale_hit' options. The first to reduce possibly large full object requests down to the smaller 304's. The latter to reduce the amount of 304's to one at a time. Amos > > Thanks for any help again. > > >>?Use refresh_pattern entries to override the max-age. > > >>?On Fri, May 02, 2008, Paul-Kenji Cahier wrote: >>>?Hello, > >>>?In our current situation, we are trying to have "Cache-control: >>> max-age=0" headers from clients to be ignored >>>?in the cache decision process, while keeping all of the 'Cache-control: >>> no-cache' and 'Pragma: no-cache' >>>?still valid as making revalidation mandatory. > >>>?Without trying to do anything, when squid receives the max-age=0 >>> directive, it decides to TCP_REFRESH_HIT since >>>?the client asks it. > >>>?Our current approach was the following: >>>?acl static_content req_header Cache-control max.age=0 >>>?header_access Cache-Control deny static_content > >>>?While the acl is properly matched, it seems the header_access does not >>> ever get applied when deciding of what to do, >>>?with the result that it's effectively being ignored. > >>>?Is there any way to make it be applied earlier/another way to ignore >>> only 'Cache-control: max.age=0' headers? >>>?(we would also preferably rather be able to define that with an acl so >>> we can only apply that directive to >>>?really probably static content) > >>>?The whole goal is to avoid firefox's F5/refresh button from forcing >>> thousands of TCP_REFRESH_HIT/304 all the time, >>>?which not only strains the servers but takes longer. Of course we also >>> want users that want to force a refresh >>>?(through ctrl+shift+R, which actually adds the no-cache directives) to >>> be able to do so.(Caching is good, >>>?but forcing delays before things are checked again is not) > >>>?Any suggestions will be really appreciated... We have tried to rewrite >>> urls through privoxy, but it came messy >>>?and fairly heavy on load, so a squid only solution would really be >>> best. > >>>?-- >>>?Best regards, >>>? Paul-Kenji Cahier >>>?mailto:pkc@xxxxxxxxxxxx > >
