Re: [Isis-wg] Request for WG adoption of draft-xu-isis-mpls-elc-00

[Robert Raszuk <[e-mail]>]

Xu,

> Since he MPLS-based SPRING or SR architecture has no change to the
> MPLS data-plane, why can't we apply the EL concept directly to SPRING
> or SR architecture?

You sure can. That was not my point. My point was that we should
analyze and also consider alternatives other then EL approach.


> be orders of magnitude less in the SR concept) is true. Compare LDP with
> the IGP-based label distribution protocol, why do you believe the latter
> means orders of magnitude less of bindings?


Because as it has been already proven in other threads number of SR
nodes in a given domain may be much less then number of IGP nodes. (I
remove the adj SIDs requirement for now). In the latter and in the LDP
based FECs you rather must support signalling of ranges of FECs end to
end for it to work. With SR you can just do it for 10 nodes in 500
node IGP domain to get pretty good results. That's key difference.


> What' the same mistakes in your mind? The entropy label concept as defined in RFC6790? Or allocate a single label rather than a label range for a given FEC?

Mistake that you need a new level of stack to support load balancing.
If MPLS would provide few additional bits in the header day one
Kireeti would not need to invent entropy labels at all. :)

With SID range we can accomplish the same without need to touch MPLS
data plane so it is fully backwords compatible too.

Cheers,
R.