The MPLS WG Archive

[Eric Gray <EGray@zaffire.com>]

Alan,

	Been a while (one or two weeks) since I looked
at that.  Possibly there is some ambiguity (I am not
trying to say that CR-LDP has the answer), but the
next phrase (fter the one that says you have to lose
the LSP-ID Hop) says that you forward the Request
message on a session associated with that (CR-)LSP.
This makes it considerably less ambiguous than if -
say - you forward a signaling message based on a 
routing look-up for the IPv4 prefix specified as the
next ER Hop in the ERO.

	I believe that what Kireeti suggest will work
in many implementations since - as he says - the IP
address uniquely identifies the LSP.  However, it is
an IP address and therefore likely to match a route
table entry.  If the implementation injects the IP
address of the LSP as a host route into the routing
table (I don't recall if this behavior is specified),
then the LSP will be the longest match.  Then, as
long as there are not two host routes to that same IP 
address, there should be no ambiguity.

	But this fits snugly under the heading of a "neat
trick" and it would be clearer and less likely to cause
trouble down the road if some mechanism were used to
specify an LSP explicitly.

> -----Original Message-----
> From: Kullberg, Alan [mailto:akullber@netplane.com]
> Sent: Wednesday, August 16, 2000 10:42 AM
> To: 'Eric Gray'; 'Kireeti Kompella'
> Cc: mpls@UU.NET
> Subject: RE: generalized MPLS signaling question
> 
> 
> Eric,
> 
> Unfortunately, the LSPID ER Hop type from CR-LDP doesn't
> help in this case because the LSPID ER Hop is removed
> before sending the REQUEST message.  See 4.7.4/crldp-03.
> 
> Alan
> 
> > -----Original Message-----
> > From: Eric Gray [mailto:EGray@zaffire.com]
> > Sent: Wednesday, August 16, 2000 12:57 PM
> > To: 'Kireeti Kompella'
> > Cc: Eric Gray; mpls@UU.NET; akullber@netplane.com
> > Subject: RE: generalized MPLS signaling question
> > 
> > 
> > Kireeti,
> > 
> > 	Hi.
> > 
> > 	I'm afraid that this doesn't help, too much,
> > in understanding how this might be done.  For one
> > thing, if the interface ID sub-object type you've
> > defined in the unnumbered draft (I am assuming you
> > refer to draft-kompella-mpls-unnum-00) is intended
> > to be used to identify LSPs, then it should maybe
> > have been defined as a larger than 16-bit field.
> > Since LSPs may be extended to large numbers of 
> > LSRs either directly or indirectly connected, the
> > number of interface IDs needed may be arbitrarily 
> > large.
> > 
> > 	For another, there's an issue with how an LSR
> > might interpret an IPv4 prefix in an ERO when there
> > are multiple routes to that prefix - including one
> > or more that traverse an LSP.  In fact, there is 
> > nothing to prevent having more than one LSP at any
> > LSR that extends toward an IPv4 prefix, is there?   
> > And, although it may make sense to use an LSP that 
> > directly connects the current LSR to a remote next 
> > hop associated with the IPv4 prefix specified in an 
> > ERO, if you cannot explicitly identify the LSP, the 
> > local routing engine may decide to use another route.  
> > This would appear to defeat the purpose of using the 
> > ERO.
> > 
> > 	So, I guess the question comes down to "is there
> > a reason why a user might want to force the path to
> > follow a particular LSP even when routing would prefer
> > a different path?"  I think the answer is yes, using
> > Alan's specific example as a case in point, so the
> > next question is - "how do we force this behavior?"
> > 
> > 	In CR-LDP, there is an explicit LSPID Hop type.
> > Perhaps there should be a "Session" sub-object for 
> > RSVP-TE?
> > 
> > 	Thanks!
> > 
> > > -----Original Message-----
> > > From: Kireeti Kompella [mailto:kireeti@juniper.net]
> > > Sent: Wednesday, August 16, 2000 9:16 AM
> > > To: akullber@netplane.com; mpls@UU.NET
> > > Subject: Re: generalized MPLS signaling question
> > > 
> > > 
> > > > how do LSR-A and LSR-B coordinate
> > > > the timeslot allocation so that LSR-A sends the data in the
> > > > timeslot on the correct Lambda?
> > > 
> > > Using the ERO.  Each lambda (actually, lambda-switched path) is
> > > uniquely identified, either by an IP address or by an "interface"
> > > index as defined in the unnumbered draft.
> > > 
> > > Kireeti.
> > > 
> > 
>