[prev in list] [next in list] [prev in thread] [next in thread] 

List:       mpls
Subject:    [mpls] MNA and pusing labels
From:       Joel Halpern <jmh () joelhalpern ! com>
Date:       2023-07-07 14:08:15
Message-ID: d2592d5d-bf9a-ceee-73d4-d04b03df202b () joelhalpern ! com
[Download RAW message or body]

To continue from my earlier email and from the discussion on the call 8 
days ago, let me rephrase.

First, to be clear about context.  The case I am concerned with is a MNA 
capable LSR that is pushing labels onto a received packet with a label 
stack that has at least one Hop-by-hop scope MNA substack in its visible 
stack region.  And where we are using MPLS Segment Routing.  I am not 
concerned with how this works with LDP.  I am not covering the case of 
RSVP-TE.

 From the behaviors described in 
https://datatracker.ietf.org/doc/draft-ietf-mpls-mna-hdr/ there is an 
implicit requirement to somehow cover two cases.  These should be stated 
as behavioral requirements in the draft.

(I am rephrasing the behavior from my earlier email because an 
interesting alternative was raised on the call.)

When a node is pushing labels without a hop-by-hop MNA substack onto the 
label stack, and this pushes the visible hop-by-hop MNA substack out of 
the visible range, it has to push something onto the stack to make the 
actions of that substack visible.

When a node is pushing labels that include a hop-by-hop MNA substack 
onto the label stack, it needs to do something to make sure that the 
actions from the incoming hop-by-hop MNA substack are still processed.  
This is needed since the document says that a node should only process 
the first MNA substack of each scope that it sees.

My earlier email proposed, and I still prefer, that the need above is 
addressed by copying the content of the incoming visible hop-by-hop 
substack into the label stack being pushed.  Either, depending upon the 
case, as a new substack or as part of the substack being pushed.

It was suggested on the call that another way to achieve the goal is to 
include in the new labels a single action that says "check the next 
hop-by-by MNA substack."  This would also meet the required invariant.  
It would add fewer labels to the stack.  But it would also require any 
node processing the label stack to be able, when so instructed, to look 
at the rest of the label stack beyond the visible stack depth.

Given that we have to specify something, and as far as I am concerned 
wew should only specify one way of addressing the issue, it seems that 
we need input from the WG to reach a conclusion about what should go in 
the mpls-mna-hdr draft.

Thank you,

Joel

PS: The issue of some actions which may want to modify multiple MNA 
substacks is a separate issue which should be debated separately 
starting from those who want to make a case for such actions.  it is 
separate from this.

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

[prev in list] [next in list] [prev in thread] [next in thread]