Re: Terminology [Re: Some Comments on ID/Loc Separation Proposals]

[jnc@mercury.lcs.mit.edu (Noel Chiappa)]

    > From: "Spencer Dawkins" <spencer@mcsr-labs.org>

    > So we have someone who can channel the spirit of NSRG...

Several people, actually...

    > I wonder if you could comment on my question about "dual-stack" - is a
    > dual-stacked host one thing or two, in the NSRG sense of the term
    > "stack"?

The thing is, I think part of the reason the NSRG settled on the term "stack"
(and I don't agree that Brian's take on why is complete, but I'll get to that
separately) is that it was *not* a carefully defined term, and so people
could agree that we needed to name "stacks" without coming to complete
agreement on what a "stack" was.

I'm not sure the NSRG ever did come to any kind of consensus on what a
"stack" was - but I'm pretty sure it had overtones of a collection of things
at *different* layers, including the application - as in "the complete
protocol stack".

For myself, to answer a previous question from Dave, I never could see a clear
difference between "stack" and "endpoint" (as originally defined in
http://users.exis.net/~jnc/tech/endpoints.txt).  If anyone can clearly explain
to me the difference between a "stack" and an "endpoint", I'd be interested -
but in the absence of such a clear distinction, I'll continue to treat them as
synoonyms.


So in that context, using the definition of endpoint as "a fate-sharing
region which is one end of an end-end communication", your question does not
have a definite answer, because it depends on how the "dual-stack" host is
implemented.

If the two protocol stacks are inextracably bound together (and probably to
that machine as well), then they are a single "endpoint", and thus a single
"NSRG stack". (They still might have different *names* in each protocol stack
- remember that a *name for a thing* is NOT *the thing itself*.)

If they are implemented such that either protocol stack can up and go off to
another machine, or something, then yeah they are separate endpoints/N-stacks.

	Noel