The MPLS WG Archive

[Eric Rosen <erosen@cisco.com>]

Neil> With all due respect Eric, I  think your comments as regarding MPLS as
Neil> an extension  to IP  is the root  of many  of the problems  people are
Neil> expressing with MPLS 

I would say  to the contrary that the folks who  tend to express unhappiness
are just the ones who fail to see that MPLS is simply an extension of IP.  

If you try to think of MPLS as a different, independent layer, then you will
inevitably end  up confusing yourself.   This confusion manifests  itself in
several forms:

- a perception that there seem to be a lot of layer violations

- a  perception that  it doesn't  conform to  some academic  architecture of
  inter-layer interactions 

- a feeling that an IP header is missing somewhere. 

I  think  this  confusion  tends  to  be  self-induced;  if  you  insist  on
interpreting MPLS  as an  instance of some  architecture into which  it just
doesn't fit, then you're just bound to get confused.

In the cases where LSPs are  carrying data link frames (or cells), there may
not be an IP header present, but the MPLS header is just taking its place. 

I think this  misunderstanding of MPLS can be found  both in "bellheads" and
in "IP bigots", though it comes out in different ways.