draft-ietf-ccamp-gmpls-recovey-funtional-00.txt

["Adrian Farrel" <afarrel@movaz.com>]

Jonathan, Bala,

Thanks for this draft, and well done for getting out well before SF meeting!

I wonder if this is a key draft to actively take to the ITU for their opinions?

I have some comments. Let me know if anything is unclear.

Cheers,
Adrian

==============

section 2 para 1
It would be helpful to clarify that a bidirectional LSP can utilize
unidirectional 1+1 span protection.  This comes out in section 2.1, but given
the re-use of the keyword 'unidirectional' it would be good to bring out the
difference more explicitly.

section 2 para 2
Why do you insist that M<=N? Although this may be normal and reasonable, I don't
see that there is any reason to enforce this.

section 2 para 2
The paragraph seems to fizzle out.

section 2.2, point 2, bullet 2
If the signaling procedures for 1+1 bidirectional protection are "the same as in
1+1 unidirectional" how can a service requester specify what type of protection
should be used where links are available that offer either (or both) types of
protection.

section 2.3
It might appear that the material for 1:1 protection and for extra traffic
should be split into separate sections with extra traffic building on the
functions of 1:1.  Can you add some text to explain why this is not the case?

section 2.3
I don't see anything here about setting up the low priority LSP that constitutes
the extra traffic on a protection span. Similarly, although you say,
"Pre-emption must be supported to accommodate Extra Traffic" you don't discuss
the signaling or repair of the LSP which has had its traffic displaced.

section 3.3
I think it might be useful to highlight that end-to-end M:N protection is only a
special case of shared mesh protection, and that shared mesh can equally be
applied to resource sharing on the protection paths of LSPs with distinct source
and/or destination points.

section 3.3
I think two sections need to be added.
- extra traffic (same rules as for span protection)
- "unprotected" and "protection restored" signaling to the LSPs that share the
protection resources

Tunneling
It would be helpful to consider the case where a "link" is itself an LSP with
transit nodes where resources must be managed.