Re: MPLS,IETF, etc..

    > From: Bob Braden <braden(_at_)ISI(_dot_)EDU>

    > I don't get why label swapping is any simpler than hop/hop forwarding.

At the time when tag-switching (to give it its original name - well, actually,
as someone pointed out, it was Ipsilon's product that started the trend, and
tag-switching from Cisco was in some ways a marketing response, one that got
IETF'd) was conceived, it wasn't really feasible to build hardware router
chips of the kind we see today used in boxes from Juniper, etc. You just
couldn't get enough gates into a chip. Tag-switching was enough simpler that
fully-hardware processing of packets was doable.


However, that wasn't the only rationale behind tag-switching; it also allowed
networks to see and control traffic aggregates - and in particular, control
that aggregate over a path - something which is obviously impossible with a
hop-by-hop design. This a capability still sadly lacking in all versions of IP
today. That (among other things) was something that kept MPLS going after one
principal motivator (hardware packet processing) was no longer applicable.

In particular, the ability to manage traffic aggregates turned out to be
exactly what was needed when it came time to deal with fully-optical networks
(which don't want to turn the photon stream back into electrons, and parse the
electrons back into packets, at each switch), and so now there is MP*S.


Now, of course, MPLS has been morphed into "ATM-with-variable-length-cells"
(after the original ATM failed to take over the world); it remains to be seen
if that extensive set of capabilities is really what the world (a.k.a. market)
needs.

        Noel