The MPLS WG Archive

[Bora Akyol <akyol@pluris.com>]

At 11:21 AM -0800 2/27/01, Yakov Rekhter wrote:
>Bora,
>
>>  >>  My question is as follows:
>>  >>
>>  >>  How do you do fast reroute when you have unnumbered interfaces in the
>>  >>  ERO as in the unnumbered draft? What out of band information is added
>>  >>  to what's in the ERO when we do CSPF? Is everyone doing it the same
>>  >>  way?
>>  >>
>>  >>  If we use the (newly defined) EERO then fast reroute is the same for
>>  >>  both numbered and unnumbered interfaces and it actually becomes quite
>>  >>  easy.
>>  >
>>  >Could you point me to the MPLS WG document(s) that specifies
>>  >how to perform fast re-route ?
>>  >
>>  >Yakov.
>>
>>  Well, there was the ID from George Swallow that we discussed in San
>>  Diego. And then there were some discussions on the list that I
>>  participated in 1998. Vishal Sharma et al, produced IDs on that
>>  subject as well.
>>
>>  Maybe it is time that we write a BCP on that topic. Would this fall
>>  into CCAMP or MPLS? In the last IETF, **all** of the fast
>>  reroute/recovery schemes were sent to CCAMP AFAIK.
>>
>>  I know how we implemented fast reroute and I am sure you know of at
>>  least two implementations.
>>
>>  So if there is interest on the WG, I can volunteer to put together a
>>  group of people to write such a BCP.
>>
>>  But my original question still stands,
>
>Your original question is ambiguous, as it didn't spell out
>which *specific* fast reroute scheme is used (and as you indicated
>above there is more than one such scheme).
>
>Also, are you asserting that none of the fast reroute schemes
>could work with unnumbered interfaces in the ERO as specified
>in the unnumbered draft ?
>
>Yakov.

Yakov

What I am asserting is the fact that unnumbered interfaces as 
specified in the unnumbered draft make calculation of a fast reroute 
LSP difficult even when you make assumptions about which interface 
belongs to which router. The specific fast reroute scheme that I had 
in mind was the one where you calculate a fast reroute LSP that 
avoids the immediate next hop.

Also see the point that Vach made in a separate email.

Bora