Cell Relay Archive

[dm@cl.cam.ac.uk (Derek McAuley)]

In article <3lngre$t5l@harbinger.cc.monash.edu.au>, jwb@capek.rdt.monash.edu.au (Jim Breen) writes:
> myoung@calon.com writes:
> 
> >In <3kt1j8$fbd@harbinger.cc.monash.edu.au>, jwb@capek.rdt.monash.edu.au (Jim Breen) writes:
> >>No-one (that I am aware of) in the ATMF or IETF is looking at the idea
> >>of TCP running directly over ATM VCs, because it really violates the
> >>generally accepted principles of interworking 
> 
> >What to we mean a "principle"? Is this an axiom based upon the definition
> >of a VC?  Is it a physical limitation of a VC?  What is it about the
> >physical nature of a VC, besides some voted upon axioms, which make
> >it inappropriate to run a transport layer directly on top of it?
> 
> Nothing really to do with VCs, except that not everyone has them. The 
> whole reason for have an interworking protocol, implemented on top of
> heaps of physical networks (of which ATM is just one), is to enable
> end-end carriage of ULPs such as TCP without it having to fuss about
> the underlying physical network(s). This is what I meant by "accepted 
> principles of interworking". We can do away with this, when and only when,
> all the known universe is on the one communication system.
> 

One can implement such a system of mapping TCP channels to VC channels
underneath the IP layer quite easily and maintain internetworking. The idea is
not new. It is done every day at many points of access to the net. See RFC1055
"Nonstandard for transmission of IP datagrams over serial lines: SLIP" and
RFC1144 "Compressing TCP/IP headers".

I think the technique may violate layering principles, but clearly not
internetworking ones!

Mac.