tytso(_at_)mit(_dot_)edu wrote:
From: Mark Allman <mallman(_at_)grc(_dot_)nasa(_dot_)gov>
Date: Wed, 07 Feb 2001 16:12:52 -0500
I am fairly unconvinced in the arguments made by Mr. Gao. However,
maybe a TCPng is the wrong way to look at things. A better model,
it seems to me, is the one followed by SCTP. In other words, let's
build a new transport that has semantics that are different from
TCP. As long as it is safe (i.e., follows good congestion control),
why should we care how many of these protocols are defined? After
we ensure the protocols are safe we can just let Darwinism take its
course.
Does SCTP really have different semantics from TCP? I thought it was
more of a superset.
- Ted
Ted:
It depends what you and Mark mean by semantics :)
SCTP has some different concepts than TCP.
Most noteably streams and un-ordered service. Even just
the unit of delivery being different does pose some semantical
differences (i.e. message versus byte stream). We have found
this quite a struggle has a group of us work on a sockets draft...
How do you map SCTP to the common sockets API... it is tricky and
has even caused some heated debate (right Kacheong?) :)
Regards
R
--
Randall R. Stewart
randall(_at_)stewart(_dot_)chicago(_dot_)il(_dot_)us or rrs(_at_)cisco(_dot_)com
815-342-5222 (cell) 815-477-2127 (work)