On 9/06/2016 4:02 a.m., Heiler Bemerguy wrote: > > Hum.. Amos, that store_miss would just make the object with that Vary > header to not be cached, right? It will prevent caching for any object using a Vary which matches any of the ACLs regex patterns. The patterns being ones which match the headers that are known to cause variant explosions. > > I've just tested it. But I think we need to choose what types of Vary > are valid/usable or not. Huh? > > Reading the source code since yesterday, I made a patch that seems to > fix my vary loop problem and will increase HITs a lot. > > It will only accept "accept-encoding" as a Vary statement and ignore all > the other garbage servers are using as "Vary"... > > A simple change on http.cc, not client_side.cc > I think your patch is identical to hard-coding the config: acl vary rep_header Vary -i . acl varyAE rep_header Vary -i ^accept-encoding$ store_miss deny vary !varyAE It is not reasonable to hard-code these things in a widely used generic HTTP proxy like Squid. If it can be configured do it that way, if it can't then config needs to be added. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
