On 25/06/18 05:32, Michael Pro wrote: > Yes! I realized it! > In store_id_program, I use the third-party utility "curl -sI > http://url.to/real.file.zip"; > to define Content-Length and return to squid using the "note" mechanism > in the desired ACL for response_delay_pool_access. Similarly, > using the "note" mechanism, I force the files, depending on the extension > and content in the cache, to use a specific tcp_outgoing_address > or the specific FIB. > > Could not you do this with the help of squid itself? Infinite loop : To determine *how* to contact a server for that URL, first contact a server for that URL. The only reason your test "works" at all is that curl request is *not* going through Squid. If you were to optimize it to send its traffic to Squid you should see the loop immediately. Also, there is no guarantee that a curl request to a server using curl U-A and Accept-* headers and transfer compression capabilities - has any relation to the actual response that same server will generate to a request containing the clients U-A, Accept-* header, Cookies, and transfer compression capabilities. NP: I've seen a LOT of servers which produce 404 pages or short web redirection pages when robots or CLI like curl connect to them - but the real download content when Browsers connect. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
