RE: MPLS label and LSE data models

[Xufeng Liu <[e-mail]>]

Hi Jeff,

> -----Original Message-----
> From: Jeffrey Haas [mailto:[e-mail]]
> Sent: Tuesday, June 20, 2017 4:50 PM
> To: Xufeng Liu <[e-mail]>
> Cc: Greg Mirsky <[e-mail]>; [e-mail];
> [e-mail]; [e-mail]; [e-mail]
> Subject: Re: MPLS label and LSE data models
> 
> Xufeng,
> 
> On Tue, Jun 06, 2017 at 07:33:05PM +0000, Xufeng Liu wrote:
> > From: Greg Mirsky [mailto:[e-mail]]
> >   *   yes, grouping mpls-label-stack covers LSE though I cannot see why it needs
> id, sequence identifier. I'd expect the label stack already be properly ordered;
> > [Xufeng] There are two ways to achieve the ordering: 1) Explicit sequence id, 2)
> Implicit order of the list items. Personally I feel that the explicit way is more
> clear and easier to use, but have no strong objection to the implicit way.
> 
> While I found the semantics of mpls-label-stack[id] to be clear, it does have the
> peculiar property that the ids present in the list may have gaps.  E.g. 10,20,30
> instead of 1,2,3.  And also the ambiguity of whether people's implementations
> starting counting at 0 or 1.

[Xufeng] I understand the peculiarity of the ID gaps, but I don't have a good way to eliminate it. It seems that YANG does not have a simple mechanism to specify an ordered list items allowing duplicated items, so we'd need to use the ID, but we cannot restrict the ID values to be consecutive easily. However, the implantation and the operator can ignore the gaps, and use the ID values only for the ordering purpose. Therefore, it would not matter whether the implementation starts counting at 0 or 1, or any other number.

> 
> I'm not conversant with common Yang tool suites, but it seems if the ordered-by
> user rather than the default of system, then the tooling might present the
> bottom of stack entry as the first or last node of the list rather than requiring the
> consumer to have to run a sort of the nodes based on the id number and then
> select the first node.

[Xufeng] Are you suggesting to use the position to order instead of the ID value? If so, what would be the semantics of the ID value? If we cannot get rid of the ID key, I don't know if using a separate ordering mechanism will make the usability better.

Thanks, Xufeng

> 
> -- Jeff