Re: [Asdf] Kicking off ASDF

[Michael Richardson <[e-mail]>]

Carsten Bormann <[e-mail]> wrote:
    > On 2020-08-08, at 00:07, Michael Richardson <[e-mail]> wrote:
    >>
    >> Signed PGP part
    >>
    >> Carsten Bormann <[e-mail]> wrote:
    >>> https://github.com/one-data-model/SDF
    >>
    >>> One Data Model people have a number of feature requests that we should
    >>> have a look at now.  It would be good to have an SDF 1.1 by October
    >>> that addresses some of these requests, and to keep up an approximately
    >>> bi-monthly rhythm of releases (modulo holiday times).
    >>
    >> 1) What would you call a "release" here?

    > An Internet-Draft that is specifically marked as such (often called
    > “implementation draft” in other groups).

    > The intention would be to reach a periodic milestone where OneDM people
    > and ASDF agree that a specific I-D is useful as a basis for further
    > OneDM work, e.g., by using that I-D for validating models in the OneDM
    > playground and convergence processes, and by ecosystem tool builders
    > modifying their tools to target that I-D in their conversion processes
    > (which is what enables breaking changes).

I'm not exactly thrilled about having an I-D hang around for a long time like this.

    > This could be done in a tick-tock style, with a feature I-D and then a
    > period of stabilizing that, so -02 could be SDF 1.1, -04 1.2, and so
    > on.

I'm even less thrilled about assigning version numbers to "Work-in-progress"
If we could iterate on RFCs faster, would that also work?

It seems to me that maybe we should fork/multiprocess:
  (1) take current document, SDF 1.0, kick it off as Informational<*>
     and go out to get reviews and approvals immediately.

  (2) simultaenously, start on 1.1/1.2 as you suggest, pulling in the
     review comments into that document as we go, producing a -bis
     document late next year.

<*> - this is the typical publish externally developed spec as Informational,
    then revise it within the IETF, and publish as "standard".
    But, I think that the SDF document might always be Informational :-)

I think that we could reserve a point prior to RFC publication of (1) at
which we might decide not to publish 1.0.

I big concern I have, thinking about publication, is that I think that the
subject matter is sufficiently esoteric compared to, say, BGP extensions or
IMAP extensions, that getting reviewers who aren't already involved with
OneDM will be difficult.

    >> 2) Are there important ODM member deadlines, meetings or processes that we
    >> should consider in our milestone process?

    > See above.  I am not aware of any specific deadlines.  OneDM meets
    > weekly at the moment, and I would expect us to have some personal
    > overlap that facilitates communication.  I’m not sure we have all the
    > details of making OneDM’s convergence process work in the presence of a
    > periodic SDF spec release process; I would expect the mutual benefit
    > from making that work to motivate OneDM to work on these.

okay, thank you.
Basically, you are saying that we can self-impose/self-organize deadlines.

    >> 3) Would SDF be a single document, or a base document plus extensions?

    > More on the side of a single document.  A spec language for protocol bindings might be a conceivable extension.

    >> Are there parts that we can rapidly gain consensus on and publish, with
    >> discussion about the other parts?

    > The consensus we have on SDF 1.0 is based on existing input from
    > ecosystem SDOs; it is also based on understanding that the current
    > limitations will be addressed by further work.  So I don’t think we
    > should be publishing anything quickly now.

    >>> Some github issues will now be opened by people who already have
    >>> experienced limitations of SDF 1.0 that should be addressed; of course,
    >>> other people can submit issues, too.  We can have a discussion right in
    >>> the github issues, or, for more fundamental decisions to be taken, also
    >>> here in the mailing list.  It will probably take a little until we have
    >>> pull requests based on these issues; we can discuss these then and, by
    >>> the time the discussion solidifies, we might have a WG.
    >>
    >>> I’m looking forward to SDF 1.1, and the further work of the ASDF WG-to-be!
    >>
    >> Should publishing 1.0 ASAP be a goal, or should we just do 1.1?

    > 1.0 was “published” as an Internet-Draft.  I don’t think we should aim
    > for RFC status until we have at least a handful of SDOs that have put
    > the majority of their relevant models through the process.  1.0 is
    > intended to be quickly superseded by 1.1, and that might happen a
    > couple more times until returns start to diminish.

    > This rapid iterative process may be a bit unusual for IETF.
    > I’d assume we can stabilize the SDF core at 1.3 or 1.4 enough to go for RFC.

I understand your concern about "release.0" issue, and that you wish to
validate things.  I want to point out that this is precisely the difference
between Proposed Standard (1.0) and Internet Standard (1.1).

Personally, I have been pushing back loudly against continued grade deflation
in the IETF process.   So I would personally be loath to put this into a charter.

--
Michael Richardson <[e-mail]>, Sandelman Software Works
 -= IPv6 IoT consulting =-