Re: [Lsr] RtgDir Last Call Review: draft-ietf-ospf-sr-yang

[tom petch <[e-mail]>]

From: Acee Lindem <[e-mail]>
Sent: 12 December 2023 22:25

Hi Tom,

> On Dec 11, 2023, at 7:45 AM, tom petch <[e-mail]> wrote:

Acee

top posting since most of my comments are addressed in -25 (which I have reviewed)

Renaming the YANG module is a pain but probably needs doing on the assumption that there will be a ospf-srv6-yang at some stage and that you will not want to merge the details of that into the module specified here.  Of course it may then be that just as OSPFv2 and OSPFv3 have much in common and have been merged into a single module, so SRv6 may get merged in in which case the original name would be better.  Sometimes you cannot win!

Tom Petch
I do see routing-ti-lfa progressing this week so hopefully Normative will be ok.





>
> A convenient addressee list so top posting my first thoughts on ospf-sr-yang,  I hope to find time to have a more detailed look, at least at ospfv2.
>
> I have looked at ospf-sr-yang and have some queries.
>
> Is this all flavours of SR or just some?  Most discussion I see these days relates to SRv6 I guess because SR-MPLS is mature in many respects  but think that this I-D needs to spell out the scope (like its lsr twin)

This specifies OSPF SR for the MPLS data plane. I’m considering renaming the data module to ietf-ospf-sr-mpls.yang as well.

> I note the import from sr-mpls and think it a mistake.  The routing RFC says that new protocols should have a presence container to switch the protocol on and off which sr-mpls does not do but I think that ospf-sr-yang should follow the guidelines.

We need to follow the sr-mpls model. We can’t change it in the OSPF SR model.


>
> There are mentions of vendor augmentations but no indications of what they might be and, importantly, where they would go.  Other I-D, anticipating augments, include containers explicitly for augments so that different vendors put the same information in the same place.
>
> I am used to ospfv2 and ospfv3 being derived identities from ospf which makes reference to one of the other or both simple, as ospf-yang does.  Why not here?

I’ve updated these to use derived-from() and the current path.


>
> I-D references seems to lack
> RFC8102
> "draft-ietf-rtgwg-segment-routing-ti-lfa -
> Latter needs to be Normative since a feature

I hate making the latter normative but I guess it needs to be hopefully the authors of this draft will finally bring it to completion.



>
> s.1.1 is ood

This has been removed.


>
> router-id is provided by RFC8294 so it should be imported and not be reinvented here

Okay - I have used this definition.


>
>   import ietf-ospfv3-extended-lsa {
> lacks a reference clause
>
>        leaf preference {
>           type uint8;
>           description
>             "SRMS preference TLV, value from 0 to 255.";
>
> so what?  what difference soes it make to be 0 or 255 or 42?

The description has been updated to indicate that an SR Mapping Server with a higher preference is preferred.

Thanks,
Acee



>
> Tom Petch
>
> ________________________________________
> From: Lsr <[e-mail]> on behalf of [e-mail] <[e-mail]>
> Sent: 05 December 2023 08:15
>
> Hi Acee,
>
> I've looked at the diff: the new version looks good to me. Thanks to the
> update.
>
> Regards,
>
> Julien
>
>
> On 01/12/2023 18:05, Acee Lindem wrote:
>> Hi Julien,
>>
>> Thanks much for your review. I’ve incorporated almost all of your comments  in the -23 version.
>>
>> See inline.
>>
>>> On Nov 29, 2023, at 11:03 AM, [e-mail] wrote:
>>>
>>> Hello,
>>>
>>> I have been selected as the Routing Directorate reviewer for this draft. The Routing Directorate seeks to review all routing or routing-related drafts as they pass through IETF last call and IESG review, and sometimes on special request. The purpose of the review is to provide assistance to the Routing ADs. For more information about the Routing Directorate, please see https://wiki.ietf.org/en/group/rtg/RtgDir <https://wiki.ietf.org/en/group/rtg/RtgDir>
>>>
>>> Although these comments are primarily for the use of the Routing ADs, it would be helpful if you could consider them along with any other IETF Last Call comments that you receive, and strive to resolve them through discussion or by updating the draft.
>>>
>>> Document: draft-ietf-ospf-sr-yang-22
>>> Reviewer: Julien Meuric
>>> Review Date: 2023-11-29
>>> Intended Status: Standard Tracks
>>>
>>>
>>> *Summary:*
>>>
>>> This document is basically ready for publication but has nits that should be considered prior to publication.
>>>
>>>
>>> *Comments:*
>>>
>>> - The very first paragraph of the introduction/overview section summarizes the basis of YANG, XML, JSON, data models... I believe we are now far beyond those general considerations and we could skip that paragraph.
>> Removed  - thanks.
>>
>>
>>> - In the grouping "ospfv3-lan-adj-sid-sub-tlvs" (p23), the leaf "neighbor-router-id" uses type "dotted-quad". This is consistent with RFC 8666 which specifies the associated OSPFv3 TLV, but we had a discussion about the type for router-id in the TE YANG models. The current resolution on TEAS side will be to consider a union of dotted-quad and ipv6-address. I wonder how much RTGWG would be ready to consider a superset of the existing OSPFv3 TLVs.
>> This is the OSPF Router-ID which is different from the OSPF TE Router-ID. The two should not be confused as the OSPF Router ID is simply a 32 bit unsigned integer that is typically represented in dotted quad format. It only need be unique within the OSPF Routing Domain. Conversely, the OSPF TE Router ID is a routable IPv4 or IPv6 address.
>>
>>> From RFC 2328 (which was inherited by RFC 5340):
>>
>>      Router ID
>>              A 32-bit number assigned to each router running the OSPF
>>              protocol. This number uniquely identifies the router within
>>              an Autonomous System.
>>
>>>
>>> *Nits:*
>>>
>>> - Multiple times in description: s/SR specific/SR-specific/
>> Fixed.
>>
>>
>>> - Multiple times in description: s/flag bits list/flag list/
>>> - Multiple times in description: s/flags list/flag list/
>> I changed these to either just “bits” or “flags” - the fact that it is a YANG list need not be included in  the description.
>>
>>
>>> - The description fields use a mix of "Adj sid", "adj sid", "Adj SID"... sometimes with hyphens (not to mention the full expansions). A single phrase should be chosen and used all along the module.
>> Changed them all to “Adj-SID” consistent with RFC8665.
>>
>>> - A few description starts with "The..." (e.g., in "ospfv2-extended-prefix-range-tlvs" on p 19, or v3 on p 22) while most of them don't. For consistency, it should be dropped from every brief description.
>> I removed “The “ from all the brief descriptions. I left it in two of the TLV description that were written as complete sentences.
>>
>>> - In the grouping "ospfv3-prefix-sid-sub-tlvs" (p 21 and all resulting pieces of tree): s/perfix-sid-sub-tlvs/prefix-sid-sub-tlvs/
>>> - In the same grouping, the description of the container should be "Prefix SID sub-TLV *list*." (and "Prefix SID sub-TLV." reserved for the following list element).
>> Fixed both in the module and tree (which was regenerated from tree).
>>
>>
>>> - In the container "ti-lfa" (p 25): s/Enables TI-LFA/Enable TI-LFA/ [Not wrong, but should be consistent with others.]
>> Fixed.
>>
>>> - In the same container (p 26): "s/Topology Independent Loop Free Alternate/Topology-Independent Loop-Free Alternate/
>> Fixed in this place and in another.
>>
>>> - In section 3 (p 37): s/The YANG modules [...] define/The YANG module [...] defines/
>> Fixed.
>>
>>> - In the same section: s/in the modules/in the module/
>> Fixed.
>>
>>> - In the same section: s/Module ietf-ospf-sr/The module ietf-ospf-sr/
>> Fixed.
>>
>> Thanks,
>> Acee
>>
>>
>>>
>>> Thanks,
>>>
>>> Julien
>>>
>
> _________________________________