The MPLS WG Archive

[Barry Hass <BHass@nexabit.com>]

Eduard,

In theory, I agree with you. But in this particular case, specifying
a requirement for one signalling protocol is liable to lead to a
discussion of which one it should be. And I REALLY don't want to go
there.

I think your suggested wording is fine. Thank you for translating
the original suggestion into standard draft-speak.

As an aside, and keeping in mind that I don't want to start a discussion
of the merits of the choice, I would be curious to know why the authors
of 2547 chose LDP, rather than RSVP-TE as a standard for interoperability.

Barry

> -----Original Message-----
> From: Metz, E.T. [mailto:E.T.Metz@kpn.com]
> Sent: Friday, January 12, 2001 1:44 PM
> To: Metz, E.T.; 'erosen@cisco.com'; Sasha Vainshtein; 'Shah, 
> Himanshu';
> 'Barry Hass'
> Cc: Alik; mpls@uu.net
> Subject: RE: Requirement for LDP support in the RFC2547bis draft
> 
> 
> I do not really like trading required (always interoperable), 
> for something
> less (if you're lucky)
> 
> If the requirement is to be removed anyway, then I'd suggest to make:
> - LDP RECOMMENDED / SHOULD be supported, 
> - any other method OPTIONAL / MAY be supported, 
> - and the presence of the PE-to-PE LSPs REQUIRED / MUST be there. 
> ("desirable" has no real meaning to it, "should" more or less 
> represents
> what Barry describes below as "encouraging LDP support")
> 
> cheers,
> 	Eduard
> 
> ps may "any other method" should be somewhat limited to a 
> smaller, more
> standard set of methods.
> 
> > ----------
> > From: 	Barry Hass[SMTP:BHass@nexabit.com]
> > Sent: 	vrijdag 12 januari 2001 17:41
> > To: 	Metz, E.T.; 'erosen@cisco.com'; Sasha 
> Vainshtein; 'Shah, Himanshu'
> > Cc: 	Alik; mpls@uu.net
> > Subject: 	RE: Requirement for LDP support in the RFC2547bis draft
> > 
> > I think Himanshu's suggested wording, or something close
> > to it, is a reasonable compromise.
> > 
> > "support for LDP as a mean for LSP setup is desirable for 
> the guarenteed
> > interoperability purposes. However, there is no requirement 
> that LDP 
> > actually be used, if some other method is available"
> > 
> > I read this as encouraging LDP support for interoperability
> > purposes, but allowing vendors who don't have it to still claim
> > compliance with the spec, albeit under a limited set of conditions.
> > 
> > Can people live with that?
> > 
>