Re: [netmod] 6087bis namespace recommendations

[Robert Wilton <[e-mail]>]

On 15/01/2016 15:20, Ladislav Lhotka wrote:
>> On 15 Jan 2016, at 15:49, Andy Bierman <[e-mail]> wrote:
>>
>>
>>
>> On Fri, Jan 15, 2016 at 6:19 AM, Ladislav Lhotka <[e-mail]> wrote:
>>
>>> On 15 Jan 2016, at 15:10, Andy Bierman <[e-mail]> wrote:
>>>
>>>
>>>
>>> On Fri, Jan 15, 2016 at 3:55 AM, Ladislav Lhotka <[e-mail]> wrote:
>>>
>>>> On 15 Jan 2016, at 12:49, Juergen Schoenwaelder <[e-mail]> wrote:
>>>>
>>>> On Fri, Jan 15, 2016 at 12:39:16PM +0100, Ladislav Lhotka wrote:
>>>>> Does this solve any practical problem? Modules are imported based on
>>>>> the module name and revision. On the other hand, it does create new problems:
>>>>> namespace URIs and their mappings to prefixes may be spread in many
>>>>> places in the code, and these would have to be manually edited after an
>>>>> I-D -> RFC transition.
>>>>>
>>>>> It would IMO be much better to use revision numbers rather than dates,
>>>>> and adopt a convention, e.g., that modules in the I-D stage have
>>>>> revisions 0.x that get bumped with each new revision of the I-D.
>>>>>
>>>> Lada, this is not how our current YANG 1.0 versioning and revision
>>>> rules work and we are not going to change them in YANG 1.1 either.
>>>> The rules we have do make a distinction between published modules and
>>>> modules that are unpublished.
>>> I am not proposing it. The problem I'd like to get solved is proper module revisioning already at the I-D stage so that implementations be able to distinguish one revision from another. Appending DRAFT to the namespace URI doesn't help anything.
>>>
>>>
>>>
>>> Changing the module name constantly does not help.
>>> It would be better to just keep using revision dates.
>> Some I-Ds don't do even that, for example acl-model uses the same revision date in subsequent I-D revisions. I checked that this actually violates a MUST in RFC 6087.
>>
>>
>>
>> If the YANG module does not change from one I-D to the next,
>> then the revision date does not need to change.
> 6087(bis) says:
>
>     It is acceptable to reuse the same revision statement within
>     unpublished versions (i.e., Internet-Drafts), but the revision date
>     MUST be updated to a higher value each time the Internet-Draft is re-
>     posted.
>
> I think it is a good idea.
It also think that it is a good idea.

Since an ID is effectively superseded by any new versions, I think that 
it is useful if a module defined in an ID has a revision date that 
matches the published ID, and also a reference back to the ID version 
that defines it.  At least if someone ends up implementing that module 
they can check its provenance.  Both of these properties would also be 
verifiable by idnits.

Rob