The MPLS WG Archive

[Yangguang Xu <xuyg@lucent.com>]

This is what I proposed in draft-xu-ccamp-gmpls-sig-reorg-00.txt  Make it simple
and put related pieces of information into one meaningful unit. I don't know why
we have to design something complicate intentionally.

Yangguang

Bora Akyol wrote:
> 
> Or alternately, one can add the router ID to the unnumbered ERO to
> remove any ambiguities and also simplify fast reroute.
> 
> Bora
> 
> At 4:34 PM -0500 2/27/01, Eric Gray wrote:
> >Kireeti,
> >
> >     Thanks, and thanks to Vach as well.
> >
> >     I guess the bottom line is - what Yakov said applies:
> >i.e. (paraphrasing) all self references refers to references
> >other than un-numbered interface IDs.  This might be a
> >point that would be covered in the BCP you mentioned...
> >
> >--
> >Eric Gray
> >
> >You wrote:
> >
> >>  Hi Eric,
> >>
> >>  >     Out of curiosity, how many un-numbered interfaces can
> >>  > B have that are known to it as interface 1?
> >>
> >>  One *outgoing* interface.  To clarify, in the picture below,
> >>  the LSP goes left to right, and the interface indices are
> >>  assigned by the LSR to the left.  So, the index on the AB
> >>  link was assigned by A.  B of course also could have assigned
> >>  an index to the BA link; if so, the value should be distinct
> >>  from any other index assigned by B.  So, the BA link could not
> >>  (in this instance) get an index of 1.  However, B could assign
> >>  1 to the BA link if it assigns 2 to the BC link.
> >  >
> >  >              1        1        1
> >  >          A -----> B -----> C -----> D
> >  >
> >  > Kireeti.