On 5/01/2016 10:39 p.m., Jason Haar wrote: > On 31/12/15 23:43, Amos Jeffries wrote: >> But that said; everything SG provides a current Squid can also do >> (maybe better) by itself. > Hi Amos > > Are you saying the squid acl model can support (say) 100M acl lists? The > main feature of the squidguard redirector was that it had indexed files > that allowed for rapid searching for matches - is this done within squid > now? (presumably it wasn't some time ago?). If so, is that done in > memory or via the acl files? (ala SG) - the former means a much slower > squid startup? > Yes. Squid always has been able to given enough RAM. Squid stores most ACLs in memory as Splay trees, so entries are sorted by frequency of use which is dynamically adapted over time. Regex are pre-parsed and aggregated together for reduced matching instead of re-interpreted and parsed per-request. SquidGuard is from the era when servers only had 100's MB of RAM, not tens of GB. So storing things on disk in files made sense. With OS level file caching in memory that can look like fast ACLs - but in reality it is still slower than directly accessing the listed value in RAM where the entries are stored in a format that can be quickly tested against the on-wire protocol data, not to mention the Squid<->helper protocol overheads. Amos _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
