The MPLS WG Archive

[Eric Gray <eric.gray@sandburst.com>]

Eric,

    If you mean that the use of penultimate hop pop reduces the
'look-up' requirement at the egress without increasing the
'look-up' requirements at the penultimate hop, then you are
correct.

    But it does incur some extra work for the penultimate hop -
particularly in the case where the last label (bottom of stack)
is being removed.  Without penultimate hop pop, the penultimate
hop doesn't even have to look to see if the current label is the
last one.  With PHP it has to do this and use different encap-
sulation depending on the result.  There is also implicit state
requirements associated with accepting the implicit NULL
label.  This is cheap, but it isn't free.

    The current wording in the RFC, states that there is no
new lookup required at the penultimate hop and - as you
point out below - it is correct.  It is not extraneous because
the statement is needed in order to demonstrate that the
desirability of the PHP optimization is justified.

--
Eric Gray

You wrote:

> On Mon, Apr 30, 2001 at 04:22:57PM +0530, Bhupesh Bajaj wrote:
> > hello,
> >
> > as far as i believe the penultimate hop is going to behave in same way ( as far as number of lookups is concerned )in both cases with / without using this technique.."penultimate hop popping"..
> >
> > so i believe it needn't be mentioned in RFC that with PNP one gets the benifit of just one lookup from penultimate hop side also as they have mentioned in RFC
> >
> > waiting for comments,
> >
>
> The statement may be extraneous, but it's certainly true.  THe idea
> was (I think) to point out that implicit-null causes no extra work on
> the penultimate router.  Is there major objection to pointing this
> out?
>
> eric
>
> >
> > thanks and regards,
> > bhupesh
> >
> > Get Your Private, Free E-mail from Indiatimes at http://email.indiatimes.com