Re: Comments on draft-katz-yeung-traffic-08.txt

["Naidu, Venkata" <[e-mail]>]

Sina,

-> when you have a A-B-C-D path and say A-B path goes over
-> multi-acess link, what is un-natural if we have more than
-> two routers attached on A-B or just two ?

 very unnatural. MPLS-TE as it stands now considers links
 characteristics (mostly bw) to do source routing. If you want
 to do TE without considering any reservations, how are you
 planning to signal the source computed path ? RSVP-TE as
 it stands now, is not tailored to run on non-p2p shared medias
 (neither the label distribution not label signaling are
 specified in RSVP-TE when more than 2 router are in the LAN).

 At least, tell me how are you planning to do TE (fine, without
 any QoS) once you distributed control plane topology information
 using OSPF ? How will you signal the path once you got the
 topology info for all the routers (> 2) in broadcast net ?
 please don't tell me you will do static TE, manual tunnels
 after all :)

 We are trying to specify things in routing protocols where
 there is no support in other modules. Maybe, we are trying to do
 backward engineering - first specify topology in OSPF-TE then
 research on how to computer path using CSPF then kludge TE
 (without QoS, without mapping qos bits to layer 2 headers)
 then think about how to distribute labels using signaling
 protocols then think about problems in forwarding plane etc etc
 - every thing is backward-engineered :)

--
Venkata.