On 2/15/07, skaller <[EMAIL PROTECTED]> wrote: > On Thu, 2007-02-15 at 12:32 +0100, Rhythmic Fistman wrote: > > > Not quite - the c10k problem is to do with serving thousands of clients, > > i.e. disk->network traffic. felix's considerable contribution to this has > > been > > the complete removal of the scheduler consideration*, pushing the > > issue back to the OS's > > socket/disk IO implementation and related readiness interfaces**. > > That's right. And that is why we probably need to use a single > raw socket on port 80, and re-implement the required part of > the TCP/IP protocol on top of that: potential commercial project > for Async P/L.
I'm not so sure, I think the better readiness interfaces/socket impls (implied here in OS form for netbsd/linux, http://bulk.fefe.de/scalability/) can already handle it. Which operations would you say are not scaling well, on which OSs? Is there any point fixing up anything that isn't linux ($500 slashdottable server) or windows ($700 slashdottable server)? Making and hopefully confirming the good measurements/graphs in the above link via a [portable] felix implementation would be excellent. ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys-and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Felix-language mailing list Felix-language@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/felix-language
