The MPLS WG Archive

[Thomas Morin <thomas.morin@orange-ft.com>]

IJsbrand Wijnands :
> 
> >> It looks to me that the table selection is a local policy and does not
> >> necessarily need to be the same on each LSR. You may want to use a
> >> MuRIB between 2 LSRs just to make Multicast divert from Unicast for a
> >> specific link. How these two tables are used is up to the operator to
> >> define. It does not seems to be something the draft must spell out.
> >
> > I don't think that the above is exact: to ensure no routing loop is
> > created, consistency between SPF calculated by different routers is
> > needed.
> >
> > If, for instance, a multi-topology IGP is used, then we may want to say
> > that all routers should use the same topology to lookup upstream
> > interface to LSP root.
> 
> I agree that with a multi-topology IGP it is the most obvious way to 
> build your tree. But it is still possible to only run a multi-topology 
> in a sub section of the network where you have specific needs to do so. 

True, there are topologies that will be loop free guaranteed even if MT
is not run everywhere. 

> I don't think it is up the mLDP draft to limit the scope.

I agree.


> > We could say for instance that multicast topology MUST be used if
> > multi-topology ISIS is used (or an alternate new multi-topology 
> > topology
> > ID, if we don't like reusing the one dedicated to IP multicast).
> 
> I see multi-topology IGP as different unicast tables. These tables 
> represent different topologies where you may want to pick one for mLDP 
> build trees. If we don't want to call one the multicast IGP, how do you 
> want to define this in the draft?

The draft might include some "warning" statement against
misconfiguration, but probably the one included in
draft-ietf-isis-wg-multi-topology, Section 6, is enough ("The nexthops
derived from the MT SPF MUST belong to the adjacencies conforming to the
same MT for correct forwarding. It is recommended for the administrators
to ensure consistent configuration of all routers in the domain to
prevent undesirable forwarding behavior.")

Cheers,

-Thomas



_______________________________________________
mpls mailing list
mpls@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/mpls