Re: [mpls] Update to the bullets on the first page of my review of draft-song-mpls-extension-header-09

[Haoyu Song <[e-mail]>]

Some observations on the MNA order.

For mutually independent MNAs, enforcing the order doesn't make any sense but become an unnecessary constraint.  
For some MANs, enforcing the order is also impossible (e.g., some MNA needs to be processed throughout the pipeline, and yet some others are only processed at some specific points). 

For dependent MNAs, their order must be specified when being introduced, but their execution order doesn't need to match their appearance order in a packet. So the processing order that jumps  back and forth from psd and isd is possible.  

Haoyu

-----Original Message-----
From: Tony Li <[e-mail]> On Behalf Of Tony Li
Sent: Monday, September 5, 2022 8:44 AM
To: Loa Andersson <[e-mail]>
Cc: Haoyu Song <[e-mail]>; Tianran Zhou <[e-mail]>; [e-mail]; [e-mail]; [e-mail]; [e-mail]; DetNet Chairs <[e-mail]>
Subject: Re: [mpls] Update to the bullets on the first page of my review of draft-song-mpls-extension-header-09


Loa,


> but we do need a deterministic, specified order of operations.
> 
> agreed, what I wonder is if we can have for example a processing order of NA1(psd)->NA2(isd)->N3(psd)?


Certainly.  For example, the evaluation order could be based on the numeric values assigned in the registry.


> If we can, doesn't have to be captured in requirements, framework and solution specifications?


I think that solution specification needs to give the order for the solution.

Since the framework document has no idea what ordering the solution might want to take, and there are several orderings that are possible, it seems like it is not the place of the framework document to specify the ordering.  Thus, the only thing that the framework document should say is that a solution should give us a deterministic, explicit order.

To be more specific, I propose adding the following text to the framework draft:

	A solution MUST specify a deterministic order for processing each of the Network Actions in the packet.

Comments? Tweaks?

Tony