The MPLS WG Archive

[Hummel Heinrich <Heinrich.Hummel@icn.siemens.de>]

1)For some while I also thought about separate label spaces, but not one space per hierarchy level (there are no well-defined
hierarchical levels and at least for now I don't think they are needed). Instead, I thought about  one label space which is
owned by the respective tunnel edge router, and another which is owned by a different entity like a domain-related server.
However meanwhile   I  think none of these label spaces are required. 

2) I don't like penultimate label popping. The gain is not worth the trouble to deal with two alternatives.
However, though mostly  penultimate label popping looks like a borne trouble-maker there will be solutions that will
bypass the problem, like solutions which do not depend on hierarchical label spaces. 

Heinrich 

-----Ursprüngliche Nachricht-----
Von: Karthik Muthukrishnan [mailto:mkarthik@lucent.com]
Gesendet: Freitag, 13. Juli 2001 16:16
An: anouk@ureach.com
Cc: mpls@UU.NET; chandrasekar kathirvelu; Andy Malis
Betreff: Re: Label Spaces and Uniqueness


Anouk:

I agree with your concerns.. These concerns are the basis of the draft: 

http://search.ietf.org/internet-drafts/draft-kathirvelu-hiervpn-corevpn-00.txt.

A complication is that with penultimate hop popping, every two consecutive labels starting with the first one have to be from the same space. For example, <L1, L2, L3, L4> are the labels with L1 being
the highest on the stack. L1 and L2 are from the same space [say space1] and L3 and L4 are from the same space [say space2] where space1 and space 2 may or not be equal. The choice of whether space1
and space2 are the same and the action when L2 is popped off is left to the implementation; in other words, no wordsmithing in the standard docs is necessary..

-karthik

Anouk Rocher wrote:
> 
> Eric, Arun, Ross
> 
> RFC 3031 in section 3.14 says (in the penultimate paragraph)
> 
>   "In MPLS, there is no notion of
>    having a different label space for different levels of the hierarchy;
>    when interpreting a label, the level of the label is irrelevant."
> 
> Does this mean that while doing label stacking, it is neccessary to
> allocate labels from a single label space? Isn't this a bit too
> restrictive for the cases where the inner label represents
> customer traffic transiting a provider and the outer label represents
> a path through the provider?
> 
> In other ways, I think it is appropriate in some situations for labels
> at different levels of the label-stack to be from different label spaces.
> 
> Comments?
> 
> Anouk
> 
> ________________________________________________
> Get your own "800" number
> Voicemail, fax, email, and a lot more
> http://www.ureach.com/reg/tag