The MPLS WG Archive

[Markus Jork <mjork@avici.com>]

> Eric> the problem with option 2  is that MPLS forwarding likes to obtain the
> Eric> data link header as a result of the label lookup
> 
> Alia> But when you remove the last label, it is then necessary to do some IP 
> Alia> header  processing  such  as   TTL  decrementing  or,  at  the  least,
> Alia> inheriting. 
> 
> You are correct, but that does not contradict my statement. 
> 
> It's clear  that performing  the TTL  processing when you  pop off  the last
> label requires manipulating  the IP header.  But the  MPLS specs say nothing
> about using the contents of the header to help choose the outgoing data link
> header, except  in the case  where the label  lookup indicates that  the LSR
> itself is the packet's next hop.
> 
> So  an implementation  that is  compliant with  the  existing specifications
> cannot necessarily  be relied  upon to  perform option 2  when a  single LSP
> carries both IPv4 and IPv6, and PHP is used.

Ok, but we are talking about bending the rules of existing specifications
anyway (signaling IPv4 but sending both, IPv4 and IPv6). If you want
to play tricks like that, you better have the hardware to handle it
(including correct TTL processing).

Markus