> -----Original Message----- > From: Amos Jeffries [mailto:squid3@xxxxxxxxxxxxx] <snip> > 3.0 was about parity with needs. It failed some in that regard. > 3.1 is about making up that failure plus some. > Is seamless IPv6, SSL control, and weighted round-robin not enough of a > killer app for you? > SSL control is nice, but we don't use SSL anywhere near squid, so it's not a big issue either way for us... we already have weighted round-robin in -2 by using CARP with specific weights, unless you're talking about something different? As for IPv6... eh, I suppose that'll be nice if IPv6 actually starts getting use sometime this decade. As Mark said, multicore would be quite awesome, as would better memory management, better I/O throughput on the cache_dir, proper support for memory only caches, support for acl based cache_dir's (i.e. cache_dir foo allow dstdomain blah while denying everything else and cache_dir bar allow dstdomain boo while denying everything else) to improve overall hit-rate and decrease cache file flapping, handling of a cache_dir failure that doesn't include squid dumping core, HTTP/1.1 support, options support. Things in 2.6 I'd like to see in 3 (on top of mark's list): COSS support - stable, with all the functions -2 has follow_x_forwarded_for refresh_stale_hit umask support -Tony
