The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2000-Oct> msg00158



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

Query on draft-ietf-mpls-label-encaps

  • From: Curtis Villamizar <curtis@workhorse.fictitious.org>
  • Date: Mon, 16 Oct 2000 13:59:49 -0400
  • cc: "'curtis@avici.com'" <curtis@avici.com>, Sudhanshu Jain <sjain@maplenetworks.com>, erosen@cisco.com, Yakov Rekhter <yakov@cisco.com>, dtappen@cisco.com, tli@procket.com, dino@procket.com, mpls@UU.NET


In message <9DC5E2ABE65BD54CA9088DA3194461D6010C99F1@BBY1EXM01>, Shahram Davari
 writes:
> Hi Curtis,
> 
> The egress node of an LSP, knows that it is the egress of that LSP. So the
> penultimate node could basically swap the top label with any label (not just
> 0, 2) and the egress LSR (knowing that it is egress point)could pop it. So
> my questions are:
> 
> 1) Why do we need explicit null labels?
> 2) Why do we have two explicit null labels 0 and 2?
> 
> Regards,
> -Shahram


Whether "we" need these depends on who "we" is.  Some routers cannot
forward at full line rate if they have to do a lookup to determin that
they have to POP a label and then do a second lookup on the underlying
protocol.  These routers need to send label 0 or 3 for IPv4 and 2 or 3
for IPv6.  If they would rather special case 0 and 2 in hardware and
do the appropriate lookup, they can.

I don't see that these do any harm.  Just leaving label 3 in place as
a signaling convention and dropping 0 and 2 would be OK by me.  I
don't see a compelling reason to drop them since supporting them is
almost trivial.

Curtis