On 27/10/2015 1:57 p.m., Kris Linquist wrote: > I’ve got a bunch of services connecting to third party APIs. Many of > these APIs register hits even if squid reaches out to the service to > check the HTTP headers (I verified that hits with the log > TCP_REFRESH_UNMODIFIED/200 are being counted against me with the > third party service). Refresh is more efficient than either HIT or MISS. The origin *has* been contacted, but it said the cached content was still good to be used. This is HTTP/1.1, not HTTP/1.0. Squid is a cache, not an archive or buffer. > > offline_mode on seems to never return fresh results, no matter what > my refresh_pattern is. That is pretty much what offline_mode is designed to do. Cache everything no matter how stale and serve it while the origin has possibly gone offline/down for maintenance. > I want fresh results every 15 minutes, Sqiud is designed to output fresh results. Full stop. Serving non-fresh results is dangerous unless the origin has explicitly stated that they may be served. > otherwise return HIT from memory and don’t reach out. Is this > possible? Perhaps I just haven’t found the right combination of > refresh_pattern options. > That is not how HTTP works. Caching has a fixed algorithm for freshness built into the specifications. The origin server is in control of the parameters for that. It states how long the content is fresh for and what the proxy is permitted to do with it. Only when it leaves a parameter unstated can refresh_pattern be used to supply the default value you want. As I mentioend to Yuri in another thread recently, Ths is very much akin to liceneing of copyright materials, where the origin is the content owner and asigning rights to redistributors (proxies). Taking on other rights that have been explicitly denied is not a good situation. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
