Re: [netmod] versioning procedures (RFC vs. I-D)

["Joe Clarke (jclarke)" <[e-mail]>]

On Apr 2, 2020, at 12:01, Andy Bierman <[e-mail]<mailto:[e-mail]>> wrote:

Hi,

I agree that a revision-label could be useful in an I-D but not to indicate NBC changes (because it doesn't).
The rules need to be clear and simple with no exceptions.

 1) Special version 0.x.y contains NO NBC information
     Major version = 0 means the module has no published version

 2) First published version is 1.0.0

 3) The revision-label in an unpublished module has a special form which simply identifies
      the source of the development and the iteration of the work-in-progress.
      You can't really pick the next published label until the module is ready.

>From my example:

draft-00:   0.1.0
draft-01:   0.2.0
draft-02:   0.3.0
RFC-1:    1.0.0
bis-draft-00:   1.0.0+1
bis-draft-01:   1.0.0+2
bis-draft-02:   1.0.0+3
[repeat NBC step bis-draft-02 10 times]  1.0.0+4 .. 1.0.0+13
RFC-2:  2.0.0   (in general: 1.0.1 or 1.1.0 or 2.0.0)

The BC vs. NBC distinction is not relevant for a work-in-progress.
We have seen many times in this WG where a NBC change was made
and then later undone.  There is no value in tracking the module during development.

I support this scheme.  Rob raised it on the call today as well.  Since YANG semver is a superset of semver 2.0.0, these pre-release and build appendixes will work and could be a way to handle bis drafts well.

Joe