Re: WG LAST CALL on draft-ietf-ppvpn-rfc2547bis-03.txt

At 01:21 PM 5/19/2003 -0400, Eric Rosen wrote:

>Mark> Perhaps a  more specific name  should be applied.  "BGP-MPLS  L3 VPNs"
>Mark> perhaps?
>
>I'll probably go  with "BGP/MPLS IP VPNs", to head  off questions like "does
>it work with IPX and Appletalk".

sounds fine to me


>Mark> Does the scheme work if routes  to the same subnet are advertised from
>Mark> VRFs in 2 PEs, using different RDs?
>
>Yes.  Obviously  though only one  of those routes  will be installed  in the
>VRF.
>
>Mark> I don't think the current memo  says for this case how the PE learning
>Mark> these routes should decide which to install.
>
>Correct, it does not.  This question comes up regularly.  One could specify,
>for example,  that the one  (or more, if  load balancing is  being provided)
>installed should be the one that BGP would have chosen if the two routes had
>had the same RD.  Or one could  specify that the one installed should be the
>one  whose  BGP  next  hop  is   the  closest.   There  may  also  be  other
>possibilities.  However,  leaving it as an implementation  decision does not
>seem to cause any loops or interoperability problems, so it is probably best
>to leave it unspecified.

I would be more comfortable with that if the document at least mentioned 
that this case exists and needs to be dealt with.  It is a bit subtle and 
it is not unlikely that people might build implementations that do not 
address it.


>Mark> Isn't an LSR allowed to support only DU?  And its neighbor only DOD?
>
>Good catch, so  per section 3.5.3 of RFC 3036 we  should really require that
>DU be supported on non-LC-ATM interfaces, and DoD on LC-ATM interfaces.

fine by me.
probably the text should also mention LC-FR, in the DoD camp.