Re: [mpls] comments on draft-kompella-mpls-special-purpose-labels-02

[Jeff Wheeler <[e-mail]>]

On Thu, Mar 28, 2013 at 10:11 AM, Curtis Villamizar <[e-mail]> wrote:
> If something like GAL is proposed, then an upgraded LSR could
> potentially send a special purpose label to an older LSR that is using
> that label as a forwarding label.

I don't agree.  An SP label shouldn't end up at top-of-stack unless
the receiving LSR has signalled support for that extension.

The problem is not top-of-stack issues, but middle-of-stack.  If PE2
is legacy and assigns label 17 for an application label, but P1, P2,
etc. are aware that is a new SP label that has some defined purpose,
they may take action on that packet based on the specification of SP
17.

All that needs to happen here is 0..15 be replaced with 0..N and
enough time to pass before the 0..15 space is exhausted that you can
reasonably expect all PEs in a network to receive software upgrades
before any features are specified and deployed to P nodes using SP
values > 15.  This is a very reasonable expectation given the likely
huge lead-time before any such extended SP labels are required to be
allocated (if ever.)

The question is, which of these is more likely:
1) networks depend on PE that can't receive this software upgrade in
many years time; or
2) networks depend on P that can't gain hardware ability to process
deeper label stack in many years time

In both cases, the time-frame is years; but #2 is where hardware
limitations on label stack depth arise.  Folks in this WG have argued
that hardware limits are an issue for changing 0..15 to 0..N and there
is no evidence of that.  There is a large body of evidence that many
routers have degraded or unspecified behavior once label stack grows
beyond a certain depth, and these routers continue to be sold into
customer networks today.  I do not expect that will change, or grow by
2 labels, because of this draft.  Do you?

Quite simply, to have extended SP labels based on my suggestion, all a
customer must do is request his vendor produce software updates to not
use 16..N as forwarding labels.  To have them based on this draft, he
may need to replace all of his hardware.  I think we can all agree
that software updates are more realistic.

-- 
Jeff S Wheeler <[e-mail]>
Sr Network Operator  /  Innovative Network Concepts