RE: MPLS,IETF, etc..

[Harald Tveit Alvestrand <harald at alvestrand.no>]

--On 4. september 2001 02:32 -0400 "Natale, Robert C (Bob)" 
<bnatale at lucent.com> wrote:

> I do believe that the MPLS -> MPlS -> G-MPLS expansion to
> accommodate PSC and TDM -> LSC -> FSC devices is a beneficial
> and natural extension.  The benefits offered by the traffic
> engineering opportunities would be hard to pass up too.

my personal (and largely irrelevant) belief about this evolution is that we 
are reusing mechanisms (GMPLS, MPLS-TE, RSVP-TE, CR-LDP) that were 
originally (?) evolved for MPLS tunnels to control other kinds of paths.

this is probably good (commonality is good).

However, I do not see how it validates tag-switching or the original MPLS 
design.

To me, there is a fundamental difference between two classes of decisions:

- decisions taken at setup time (lambda allocation, routing computation, 
path establishment, Strowger switch contact movement)
- decisions taken at packet switching time (IPv4 packet routing, ATM cell 
switching, MPLS label switching)

I have heard argued that MPLS (the tag switching thing) was irrelevant on 
the day that it was economically feasible to deploy 32-bit-wide associative 
memory for routing table lookup in hardware. AND that the chief benefit of 
the 6to4 and other 6-in-4 wrapper mechanisms is that you don't have to 
develop 128-bit-wide associative memory for core switches (yet).

And arguments (like Bob's) that the benefits are real and present.

But I haven't yet figured out whether the latter class of argument applies 
to the setup protocols or to the actual tag switching.
I'm sure I'll figure it out (sooner or later).

                      Harald