The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2006-Jul> msg00061



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

[mpls] context labels

  • From: Eric Rosen <erosen@cisco.com>
  • Date: Tue, 18 Jul 2006 12:15:57 -0400
  • Authentication-Results: rtp-dkim-2.cisco.com; header.From=erosen@cisco.com;dkim=pass ( sig from cisco.com verified; );
  • Cc: mpls@ietf.org
  • DKIM-Signature: a=rsa-sha1; q=dns; l=1798; t=1153239358; x=1154103358;c=relaxed/simple; s=rtpdkim2001;h=Content-Type:From:Subject:Content-Transfer-Encoding:MIME-Version;d=cisco.com; i=erosen@cisco.com;z=From:Eric=20Rosen=20<erosen@cisco.com>|Subject:Re=3A=20[mpls]=20context=20labels=20|To:Loa=20Andersson=20<loa@pi.se>;X=v=3Dcisco.com=3B=20h=3DvyjaOPLBszCWPYTG7sU6jWK15cQ=3D;b=0R18H3t68ICj3DSbWwdGIWFLGGEKpW+QvDgTl0EQQ2iQRiGPhz4Zjn6FIBkDnv5Azh8XMrtoBepYlMx0YqUooDgFzBhQRFOTzBQ+iz7negXVHlBkT+RnFO0QY3VVKgRU;
  • X-IronPort-AV: i="4.06,255,1149480000"; d="scan'208"; a="93081856:sNHT23488376"


> a question on the context label, where would it go? 

It has to be  above (i.e., closer to the top of the  stack) the UA label for
which it provides  the context.  On an ethernet,  if the ethertype indicates
"UA label",  the context label  would be  expected to be  at the top  of the
stack,   with   the  UA   label   beneath  it.    On   a   MP2MP  LSP,   the
downstream-assigned label assigned by mLDP  would be at the top, and beneath
it would be the context label and then the UA label.

> Normally when a label is propped you can forget it, because the next label
> in the stack is unique

Well, not necessarily.   If it is known, for example, that  when a packet is
received  from interface  X its  top  label is  looked up  in an  interface-
specific label space  determined by X, then the label  is not "unique", it's
only unique relative to X.  We've  always allowed the scope of uniqueness to
be determined by the incoming interface. 

If we allow the  "incoming interface" to be the context in  which a label is
interpreted, why  not generalize  and allow an  LSP egress to  use "incoming
LSP" as the context?  

When using UA  labels over a P2MP LSP,  we don't use PHP on that  LSP, so at
an LSP egress of the P2MP  LSP, the top label identifies the LSP (implicitly
identifying the  transmitter), and  hence provides the  context in  which to
interpret the next label, which is the UA label. 

It's just  a short step  to the generalized  notion of context label  that I
proposed in Montreal.

> This strikes me as being different from the normal behavior of an LSR

Think of it as a way of providing a 40 or 60 bit label space ;-)

This  is  a  change  to  the  forwarding path  and  existing  LSRs  may  not
necessarily be able to handle it. 







_______________________________________________
mpls mailing list
mpls@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/mpls