The MPLS WG Archive

[Kireeti Kompella <kireeti@juniper.net>]

> The OIF's use is hardly compelling. Curtis makes a good
> observation that the word "open" isn't particularly descriptive
> in this context.

I would go further: "open" is a misnomer.  As Jonathan pointed out,
the "open" model looks closed from another point of view.  Note that
the terms "peer", "overlay" and "integrated" all describe topological
relationships between network elements, but "open" instead describes
services offered.  Non sequitur.

> Are you *sure* you aren't really talking about
> a form of overlay model (as clarified)? The word "overlay" (with
> or without "signaled" as Curtis suggested) seems to be more
> descriptive (naturally, IMHO).

IMHO, here's what is going on: the term "overlay" may appear to some to
have bad connotations; the term "open" has good connotations.  So,
while it is nice to have descriptive names, let's call this model X
(and, if desired, call the peer model Y) to lose the connotational
baggage, and get on with the real task of defining the models by
property and context of usefulness.

Krishna, you make the comment:

> It is unclear at this time which of these models will prevail.

I don't know that any one model will "prevail".  Furthermore, by
designing a routing and signaling architecture that encompasses all
models, it won't matter which prevails, and will also allow for a
smoother transition from one to another.  I would say our real
mission here is to arrive at such an architecture ....

Kireeti.