[c-nsp] IP LFA in ring topology

[Saku Ytti]

On (2012-11-29 15:24 +0100), Adam Vitkovsky wrote:

> But what if R3 still sees the best path towards P1 via R5 direction (leg of
> the ring)
> R5 would have to impose a label stack to get all the way to R1 or basically
> to a node on the ring that can be used to forward the traffic to the
> destination loop free

You only have to impose minimal stack, as told by SPF.

But I gave it bit more thought with rLFA with regular labels and with
global labels. And I'm not anymore sure if my idea is any better, and
certainly my idea is not deployable, as no global labels exist. rLFA you
can deploy today, I'm impressed.

My idea would also fail, if there are some links which are 100% backups,
not leading to anywhere but one direction. 

-- 
  ++ytti