Cell Relay Archive

[ronald@tezcat.com (ronald davis)]

i read the ietf internet draft on tag switching.  then i compared it to
the integrated pnni description from the atm forum.  as far as i can tell
"tag switching" is basically i-pnni with router-centric peer groups and
with a priori svc's between routers (or "tag edge routers" to use the
terminology of tag switching).

the main contribution made by the tag switching internet draft was in
more clearly defining *how* the svc's are to be set up: e.g. destination
based routing, hierarchical routing, multicast routing, &c.  tag switching
also describes a framework for multiprotocol operation (something which was
mentioned as a possibility in the i-pnni description but not discussed
further).

tag switching, like i-pnni seemed to kind of handwave on the qos issue.
for instance, the people advancing tag switching say that it is better
than ip switching in that it avoids the "label proliferation" that
purportedly results due to the fact that ip switching associates a label
with each end-to-end flow.  by contrast, so the reasoning goes, tag
switching associates a tag with a route so that all "flows" using the
same link can use the same tag.

ip switching establishes flow tags as the svc is established; tag switching
avoids the latency associated with svc establishment by establishing all
svc's a priori (as stated above).  this would seem (to me) to make tag
switching subject to "route proliferation"; tag switching would have to
establish routes between all tag edge routers for all possible qos' *in
advance*, whether they actually end up getting used or not.

i would like see what others have to say about my comments.