Re: Juniper as egress LSR: to penultimate hop pop, or not?

From: Eric Osborne (eosborne@cisco.com)
Date: Mon Apr 02 2001 - 23:50:46 EDT


On Mon, Apr 02, 2001 at 05:28:24PM -0700, Serge Maskalik wrote:
>
> Eric, that's not what Lane is asking. He wants to know if
> an option exists on the Junipers to do ultimate hop popping.

Right; I was responding to Jesper's assertion that there's no way to
do implicit/explicit null on cisco gear.

> The feature you described was put in because the default
> IOS behavior is to signal label value 0 as "implicit null",
> which does not correspond to spec.

NOTE: anyone who's offended by cisco-specific email in their
juniper-nsp inbox, feel free to note the crossposting on this thread
and then delete it. :)

Default IOS behavior is to advertise 0, which is explicit null.

With 'mpls traffic-eng signalling advertise implicit-null' configured,
a TE tunnel tail will advertise 3, which is the correct implicit-null
value per rfc3032.

There's additional knobs that need to be turned to convince a
penultimate hop to interpret the 0 as explicit, rather than converting
to implicit. I wasn't doing MPLS stuff when that all went down, so I
don't understand enough of the history to get into why it works that
way. But the issue is with the interpretation of the advertised
value, rather than the advertised value itself.

> Lane, on both Juniper and Cisco, you can't change the
> penultimate hop popping behavior.

Not true, at least not on cisco stuff. We advertise 0 by default and
can change to 3 with 'mpls traffic-eng signalling advertise
implicit-null'. Easy enough to veryify - flap this knob on the tail
end and check 'sh mpls traffic-eng tunnels' or 'deb ip rsvp resv det',
if you're really curious.

Anyone who's interested in the knob to change interpretation of
explicit-null on the RESV-receiving (PHP) side, send me email and
we'll go over it offline.

eric



This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:41 EDT