fwiw, I fully agree with kre Brian
Robert Elz wrote: > > Date: Wed, 17 Jul 2002 16:13:18 +0900 > From: [EMAIL PROTECTED] > Message-ID: <[EMAIL PROTECTED]> > > | i don't think we need to un-tie flowlabels and sockets, at least > | within the default behavior. it complicate things too much (and if we > | add a new system call, we'll need to go through POSIX/XNET/whatever, > | which will be a lot of pain). > > Oh sorry, I obviously wasn't very clear. That degree of breaking the > relationship wasn't what I intended. Using setsockopt() or similar > will be just fine. All I meant was that there shouldn't be a one flow > label to one socket assumption anywhere. Several sockets might be > using the same flow label (and not just ones in the unix world split > by dup()/fork() semantics) and one socket might use different labels > at different times (within the same transport level connection). > > | for normal usage 1:1 relationship between socket and flowlabel should > | be ok. > > Yes, most of the time, for most apps, that will be fine. But we do > need to provide for the exceptional cases too. > > kre -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
