In some off-list mail between Mike and I, he said: >> Was TCP a bad idea because it didn't have MTI port numbers? Would >> it have improved TCP to add an MTI port or two?
To which I responded: > Ports are MTI for TCP. [1] They are 16 bit values > with a well-defined test for equality and a little bit > more structure/IANA stuff. We agreed it'd be useful to bring this to the list since it maybe captures the disconnect. I'm not sure I quite get it but I think Mike means that no particular port number (or the associated protocol) needs to be listened for by all TCP stacks. That's correct, but nothing to do with TCP interop. Port numbers do need to be specified by all TCP packets or those are malformed and all TCP stacks need to know how to compare those and probably more subtleties but support for port numbers is definitely not optional - its mandatory. Cheers, S. [1] http://tools.ietf.org/html/rfc793#section-2.7 _______________________________________________ OAuth mailing list OAuth@ietf.org https://www.ietf.org/mailman/listinfo/oauth
