a few comments on TLI's

[Tony Li <[e-mail]>]

   1. why do we have to upgrade all routers when a new policy is added?
   just route servers that compute routes, and just BRs or PGs of the
   domain that needs to put new policy info in its update...

Sorry, I was unclear.  I think that you have to change all IDPR
speakers, yes?  At least everyone who is going to parse policy terms.

   2. I dont see the consistancy problem that you rae alluding to, but
   need to discuss it more.

Let's suppose that I would like to construct a mini-model of the
unified scheme.  I would like to run BGP and IDPR in the same box.
What does the FIB look like?  I _suspect_ that you have to hand all
packets to IDPR.  Then somehow (magic here ;-) IDPR can pass some of
them off to normal forwarding.

   3. I hear you re. the risk of leading to deployment of an immature
   protocol and am leaning more and more in that direction. Can someone
   who disagrees convince me otherwise?  My one question re. what your
   claim re. going to experimental protocol is  that
   I dont see VMTP as something thatkcreated a problem in this sense so
   is it really the cas ethat experimetnal protocol status introduces a
   problem?  

I don't think that the analogy holds.  VMTP was and is not the only
protocol that addresses its problem space.  If you want another closer
example, look at STII.  Yes, that is also creating a problem.  My
feelings are that a either not-recommended or informational is more
appropriate to remedy this problem.

Tony