<Note the cross-posting to other groups>
Dan
For hierarchical LSPs, there are instances where you would like to
have the ability to do double lookups. It also is necessary if you
want to preserve the EXP bits for the ultimate HOP to see them. It
again becomes necessary if you are doing L2VPNs and carrying packets
other than IP.
I think that PHP is an architectural mistake made to cover for the
inability of routers to do the double lookups (a.k.a. hack).
Bora
At 7:40 AM -0400 4/3/01, dan martin wrote:
>its not a won't, its a shouldn't.
>
>i think the standardards goal was to keep a single router on the edge from
>doing two operations on the same packet, popping the label and forwarding
>the ip frame.
>
>i think asking a router to process a packet twice is the equivalent of the
>double sided copier, a simple design idea that leads to a tortuously long
>paper path.
>
>-----Original Message-----
>From: Lane Patterson [mailto:lpatterson@equinix.com]
>Sent: Monday, April 02, 2001 4:43 PM
>To: 'juniper-nsp@puck.nether.net'; 'mpls-ops@mplsrc.com'
>Subject: Juniper as egress LSR: to penultimate hop pop, or not?
>
>
>
>It appears that if I want to use a Juniper as an egress LSR, I MUST
>use penultimate hop popping? For example, using signalled LSPs,
>juniper will only send RESV's with a "3" label (implicit NULL), and
>I don't see knobs for it to use an IPv4 explicit NULL "0" label.
>
>Is this because the juniper forwarding path is incapable of doing
>both a label pop and an IPv4 forwarding decision at line rate on
>the egress LSR?
>
>Cheers,
>-Lane
>
>Lane Patterson <lane@equinix.com>
>Member Research Staff
>Equinix, Inc.
>(650) 316-6012
This archive was generated by hypermail 2b29 : Mon Aug 05 2002 - 10:42:41 EDT