The MPLS WG Archive

["Abes, Andi" <aabes@quarrytech.com>]

Hi again,


thanx to Shahram Davari for giving me the reference -

The MPLS arch allows (section 3.25.2 - bullet 2. SVP Encoding)
to encode 2 labels in the VPI/VCI fields, and states
that this might not always be usable.

So, when allocating a higher label (not a top one),
it could concievably by used in an SVP encoding - 
like in the case of 2 directly connected LSR using 
LCATM interfaces.

In this case, the second label has to be from a per-interface
label pool.

Is this encoding still supported ?
Any of this make sense ?


Can a statement that says that higher level label
are always from the per-platform label space still 
be made ?


Andi.



> -----Original Message-----
> From: Abes, Andi [mailto:aabes@quarrytech.com]
> Sent: Thursday, April 06, 2000 11:33 AM
> To: MPLS mailing list
> Subject: Higher labels and RSVP
> 
> 
> Hi, 
> 
> I'm trying to resolve a question that looked simple at first:
> 
> Is it correct to say that higher level labels can only be 
> drawn from the 
> per-platform label space ?
> 
> At first it seemed natural, and then I even found in my 
> archives discussions
> before the BGP draft was issued that reinforced my conviction:
> 
> since packets labeld with BGP labels can arrive at an LSR 
> over different 
> interfaces, and the route my vary depending on IGP activity,
> at packet reception time - the LSR can not tell by what interface the
> labeled
> packet should have arrived at.
> 
> 
> This is all fine, and sound fine for BGP assigned labels. But...
> When labels are assigned using RSVP the whole label stack 
> is assigned at once - such that it depends on the semantics  -
> is the label stack valid only as a whole ? 
> Can the LSR that received the label binding (with 2+ labels)
> use the higher levels stacked on other top labels ?
> 
> If the former is the case - then per-interface label space 
> can be used.
> if the latter, then only per-platform space can be used.
> 
> To make things interesting, in draft-swallow-rsvp-bypass-label-00 it
> is explicitly assumed that the higher labels can be used with 
> different
> top labels (one for the original tunnel and a different one 
> for the bypass
> tunnel).
> 
> Ok,
> so what say you ?
> 
> 
> Andi.
> 
> 
> 
> 
> 
>