Re: [Idr] Progressing draft-bhatia-bgp-multiple-next-hops-01.txt

[Curtis Villamizar <curtis at occnc.com>]

In message <92c950310608310720n1939b3a8p6701d4435e5c3366 at mail.gmail.com>
"Glen Kent" writes:
>  
> A quick Q while we're discussing various mechanisms to advertise
> multiple paths.
>  
> What is a router supposed to do if it recieves two paths from a BGP
> speaker as described below?
>  
> Path 1 - NLRI 192/8 AS_PATH {10 20} LOCAL_PREF 100 NEXT_HOP 1.1.1.1
>  
> Path 2 - NLRI 192/8 AS_PATH {10 20} LOCAL_PREF 189 NEXT_HOP 1.1.1.1

If this came from the same router then complain because the sender of
this is broken.

On the priciple of "being liberal in what you accept" just toss the
second path.

> What useful information can any router possibly derive from this?
> Which route should it select? I see now value and would appreciate if
> someone can enlighten me on this.
>  
> Another day, another time, but the same BGP speakers.
>  
> This router recieves two paths for 192/8 from the same BGP speaker.
>  
> Path 1 - NLRI 192/8 AS_PATH {10 20} NEXT_HOP 1.1.1.1
>  
> Path 2 - NLRI 192/8 AS_PATH {30 20} NEXT_HOP 1.1.1.1
>  
> How does this router determine which route is being used by the
> upstream? I dont think we should progress any proposal without
> incorporating ways to indicate which path is being actually used in
> the forwarding.

If its a multipath it is considering both to be valid.  This may be
possible at an IX if next-hop-self is not set and there is an odd set
of routing policies.  btw- no where in BGP does it require that
routing policies be sane (and they can be seemingly quite wierd).  If
the AS are 65010, 65020, and 65030, this is perfectly valid in an AS
confederation setup.  In AS confederations you'd just forward to
1.1.1.1 but retain the AS paths for readvertising.

> I know one of the authors of multiple_hop indicated that this could be
> easily done by introducing a field in the path attribute. Its not
> difficult but they need to show this to us. The same goes for
> add-paths.

With add-paths the path-id would be used to disambiguate path 1 and
path 2 in the second case so I'm not sure what your point is.

> Curtis said:
>  
> > I don't recall seeing any objections to the use of path identifier in
> > draft-walton-bgp-add-paths that held up.
>  
> I think its well known that silence means lack of interest, not
> acceptance. You of all the persons, should be well aware of this.
>  
> Glen

There seems to be equal or more people arguing in favor of add-paths.
Enke Chen, John Scuder, Robert Raszuk, Jeffrey Haas seem interested.
I'm interested.  I'm less sure about others.  The contributors to this
thread are (alphabetic by first name):

  Curtis Villamizar
  Enke Chen
  Glen Kent
  Jeffrey Haas
  Joel M. Halpern
  John G. Scudder
  Mach Chen
  Manav Bhatia
  Paul Jakma
  Robert Raszuk
  Tulip Rasputin
  Yakov Rekhter

To me it seems comments show interest in multiple-next-hops and
add-path even though the majority of the mail carried
multiple-next-hops in the subject line.

Curtis

_______________________________________________
Idr mailing list
Idr at ietf.org
https://www1.ietf.org/mailman/listinfo/idr