The MPLS WG Archive

["Mayank Kumar" <mkumar@aplion.stpn.soft.net>]

Hi
ur argument makes sense. Thank you . But i do not understand what do you
mean by VCID's. Is it some combination of VPI and VCI . The CISCO shows some
vcid kind of thing in its 'show atm vc' command. When u say that VCID's are
allocated on aper interface basis , do u mean that vcid's are unique for a
specific interface only and thus there can be two identical VCID's on two
different atm interfaces or for that matter ports. Then it makes sense to
not assign atm vpi/vci (labels) from a platform wide label space.
Also then i can infer one more thing that, the platform wide label space
contains only shim-labels (that means the generic encapsulation specified in
rfc 3032). Also  the pw wide label space contains just the 20 bit labels .
It does not make sense that the label space contain the 3 bit EXP and the S
and the TTl fields too.

regds
Mayank




-----Original Message-----
From: owner-mpls@UU.NET [mailto:owner-mpls@UU.NET]On Behalf Of David
Charlap
Sent: Monday, July 08, 2002 8:29 PM
To: IETF MPLS List
Subject: Re: do we need to have disjoint label space for cr-ldp and
rsvp-te


Mayank Kumar wrote:
> but then the answer to my second ques, does not give me a good reason. Why
> is it necessary to have interface specific labels for atm or frame
relay???
> how does it matter that , we have the label encapuslated in the L2 header
in
> case of atm or frame-relay.Am i missing something in saying that ,we can
> always assign a label from the platform wide label space , as long as that
> label confirms to be a valid dlci (for frame-relay) or a valid vpi/vci
> air( for atm).

The nature of ATM is to assign VCIDs on a per-interface basis.  No
existing ATM switch supports the concept of per-platform VCIDs, because
doing so would break all of the ATM protocols.

Since ATM switches, and by extensions ATM interfaces, all use
per-interface VCIDs, it makes no sense to specify per-platform labels
for these interfaces.  You'll just end up specifying something that will
never be implemented.

I assume a similar argument holds for FR.

And, as was already said, label spaces for ATM and FR are fundamentally
different from shim-label label spaces.  Shim-header label spaces use a
20-bit integer for a label.  FR uses two different size DLCIs, neither
of which is 20 bits - and there are per-interface range limits.  ATM
uses the two-dimensional VPI/VCI values, where VPI may be two different
widths, and there are per-interface range limits on each dimension.

Trying to combine all three of these incompatible technologies into one
huge pool of labels is asking for disaster.

-- David