The MPLS WG Archive

["Daniel N. Bauer" <dnb@zurich.ibm.com>]

Shahram,

thanks for your reply.

>From your answer, it follows that if the PHBs
of an E-LSP span multiple class types, then
a PathErr with code `Diff-Serv TE Error' and
some value indicating the error condition
should be sent towards the sender. Otherwise,
unexpected results may occur.

-Daniel


Shahram Davari wrote:
> 
> Hi,
> 
> > -----Original Message-----
> > From: Daniel N. Bauer [mailto:dnb@zurich.ibm.com]
> > Sent: Monday, February 19, 2001 11:20 AM
> > To: Francois Le Faucheur; mpls@UU.NET
> > Subject: Question to draft-ietf-mpls-diff-te-ext-00.txt
> >
> >
> > Hello,
> >
> > draft-ietf-mpls-diff-te-ext-00.txt describes the
> > handling of the RSVP CLASSTYPE Object.
> > Is it correct to say that for E-LSPs, the CLASSTYPE
> > object should be omitted, since a single E-LSP might
> > be associated with multiple class types?
> 
> Any operator who wants to take advantage of Diffserv TE extension, should configure its E-LSPs to support a set of PHBs that belong to a single ClassType. So in summary to support Diffserv TE extension, a single E-LSP MUST not be associated with more than one ClassType. And in case ClassType is not "0" you MUST include it in the signaling.
> 
> Yours,
> -Shahram
> 
>  But then,
> > omitting a CLASTYPE object automatically associates the
> > LSP with CLASTYPE 0?!?
> > Is omitting the correct behaviour?
> >
> > Thanks,
> >
> > -Daniel
> >