On Mon, Nov 07, 2005, Henrik Nordstrom wrote: > >Heck, I'd like to do some work on COSS which could be easily ported to > >work with squid-3 but I can't guarantee that squid-3 is > >stable/predictable enough to know whether its squid-3, or my COSS code. > > Isn't the COSS code in squid-3 quite far ahead of Squid-2 already?
I dunno, when I tried running it a while ago I had issues. I was going to cross-pollinate COSS between squid-2.x and squid-3 so we have something more useful in the squid-2 proxy I run here at work. IMHO, its (replacement FS) the last big thing thats standing in the way of excellent performance. I'm running my epoll IO changes on my work proxy. As expected, the IO stuff flatlined. I'd really appreciate some peer review just to make sure I haven't bollocks up the IO handling somehow before I move onto reworking the delay pools IO code. Adrian
