The MPLS WG Archive

["David Allan" <dallan@nortelnetworks.com>]

I get the feeling the genie is out of the bottle, and this sort of half-PHP
is really the provence of the LSR that originated the label. Only
consequence of such a spec change is the upstream LSR should not squawk or
place limitations on such a label if received.

One potential issue is figuring out who initiated the label in a PHP
scenario. The ingress may stack it unsignalled as a PID, the egress may
offer it upstream as a half-PHP. If there is PHP-LSP stacked on top of it,
technically the PH LSR is clueless as to where the label originated. So long
as you are not doing things like PHP of service labels at the egress PE, and
muxing test tools on the service label using explicit null at the ingress
PE, you should be OK ;-)

cheers
Dave

> -----Original Message-----
> From: Alia Atlas [mailto:aatlas@avici.com]
> Sent: Wednesday, January 29, 2003 10:51 AM
> To: erosen@cisco.com
> Cc: francis.arts@alcatel.be; mpls@UU.NET
> Subject: Re: Vpns vs explicit null label 
> 
> 
> I agree that removing the restriction that an explicit null 
> only occur at 
> the bottom of the stack would be a good idea.
> 
> It does raise some questions, because there is an IPv4 
> explicit null and an 
> IPv6 explicit null.  What we really want is just an explicit 
> null, which 
> means "pop off this label and then forward based on 
> whatever's underneath".
> 
> Alia
> 
> At 10:09 AM 1/29/2003 -0500, Eric Rosen wrote:
> 
> >Francis> Should the statement  that <the explicit null label 
>  only occurs at
> >Francis> the bottom of the stack> be relaxed?
> >
> >This statement causes a lot of problems, and no one can seem 
> to remember why
> >it is there.  Further, there are a number of cases in which 
> it is awkward to
> >prevent explicit  null from appearing in  mid-stack. Not 
> only  that, but the
> >specs  don't really say  what you  are supposed  to do  if 
> you  do encounter
> >explicit null above another label, or if you are instructed 
> (via LDP) to put
> >explicit null on a packet which already has a label stack.
> >
> >So I would favor removing the  requirement that explicit 
> null only appear at
> >the bottom of the stack.
> 
> 
>