The MPLS WG Archive

["Vach Kompella" <vkompella@timetra.com>]

One of the problems with the new version of the pwe3-control draft is that the
line that says that the platform label space must be used has been dropped from
the text.  I assume that this was unintentional because it doesn't make sense to
use interface-specific labels.  However, it became apparent at the ISOCORE
interop that one vendor didn't realize this.

Could it please be re-inserted, right after where it says liberal label
retention should be used?

  "VC labels MUST be allocated from the per-platform label space."

Thanks.

-Vach

> -----Original Message-----
> From: owner-mpls@UU.NET [mailto:owner-mpls@UU.NET]On Behalf Of Andrew G.
> Malis
> Sent: Tuesday, March 25, 2003 5:49 PM
> To: erosen@cisco.com
> Cc: mpls@UU.NET
> Subject: Re: Distributing Labels for PWs within an LSP TE Tunnel
>
>
> At 3/25/2003 10:04 AM -0500, Eric Rosen wrote:
>
> >Nurit> I do look for  a way to bind the (LDP) signaled  PWs to the (RSVP-TE)
> >Nurit> signaled transport tunnel LSP, in a way that both PEs know about it.
> >
> >There is no way to do this using the protocols as currently defined.
>
> I agree with Eric, and it's the way it is for a very good reason - as a
> result of requiring PW labels in the platform label space, deciding which
> RSVP-TE tunnel to use for the LDP-signaled PWs is purely a local matter at
> the PE that's encapsulating the L2 frames.  This allows a number of nice
> properties, such as the ability to very easily switch to a backup tunnel if
> there's a primary tunnel failure and the network isn't using FRR in the
> core, or to load-share over multiple tunnels if the PW doesn't have
> sequentiality requirements (or has partial sequentiality
> requirements).  Using end-to-end signaling to bind a PW to a particular
> tunnel removes these properties.
>
> Cheers,
> Andy
>
>