head hurting [was Re: [midcom] WG scope/deliverables

[Brian E Carpenter <brian at hursley.ibm.com>]

Well, I don't think this is about midcom any more but something here
made my head hurt...

Ed Gerck wrote:
...
> You miss at least one other possibility.  If it is possible to develop
> an addressing scheme that works in a heterogeneous network, then
> we can have point-to-point functionality across system borders 

er, that is what the Internet concept was invented for
by Pouzin, Cerf and Kahn in the early 1970's. The references
are in RFC 1958. That addressing scheme is called IP; the problem
is that 32 bits are no longer enough. 

> .....and
> do not require a homogeneous address space to do so.   

at some level you must have an unambiguous namespace. If 10.1.1.1 is used
in two different places there must be a way of distinguishing them.
Unfortunately, today we do this without the benefit of an explicit
namespace - the distinction is implicit in the instantaneous state
of NAT automata, i.e. the internal state of all NATs is an extension to
the IPv4 address space. 

Thus when 10.1.1.1 is behind a NAT that has loaned it address 9.1.1.1,
its implicit address is 9.1.1.1+10.1.1.1. That's an unambiguous
address space; it's just implicit. (NAPT, multiple NAT or NAT-PT make 
it a bit more complex, but don't change what I'm saying in principle -
the implicit address just gets longer.)

A rendez-vous service for NATted peers would have to construct an
identifier explicitly, and it might as well be this implicit one.

  Brian