Re: [mpls] New Version Notification for draft-kompella-mpls-rmr-00.txt

[Kireeti Kompella <[e-mail]>]

Hi Huub,

On Nov 13, 2014, at 12:59 , Huub van Helvoort <[e-mail]> wrote:

> [hh2] OK, that was not clear to me when I read the draft.

Clearly, a number of clarifications and details are needed.  Watch this space for updates :)

> [hh2] Can I conclude that the sending of messages to DS
> neighbors stops as soon as the ingress node is reached?

Yes and no.  Yes for the traffic from that node.  However, since the ring LSP is multipoint, the nodes further DS can also initiate traffic.  So, effectively, the notifications have to go all the way to the ring LSP’s anchor node.

>> In the case of the ring LSP to node K, nothing can
> > really be done, unless the real destination is
> > dual-homed (separate topic).
> 
> [hh2] OK, I would refer to that as drop-and-continue
> because there can be more than two nodes involved.

Indeed.

>>> Node L receives the message, but cannot forward it to the
>>> failed node K.
>> 
>> Not sure why you want to send the notification to node K.
> 
> [hh2] if my conclusion above is correct, then forget
> this comment.

No, you are correct in that (as I said above) every node till the ring LSP anchor needs to know, including node K.  However, one can’t just send a notification to node K — all nodes DS of the fault need to know this.

Cheers,
Kireeti.