Followup to "Supergroups" message to AC Forum

Hi Process CG,

This is followup to the "Supergroups" message sent by Jeff [1] a while
ago to the AC Forum. I was extremely busy with the release of
BlueGriffon and could not comment earlier, sorry for that. But I also
noted it did not attract a lot of comments in the list...

I ignited the discussion [2] around Supergroups almost two years ago
(two years...) with a message highlighting five recurrent issues in the
WG I know best, of course the CSS WG. I can summarize that into the
following five bullet points:

1. difficulty to give realistic ETAs, that are still mandatory in
   charters
2. binding list of deliverables
3. difficulty to add new deliverables
4. arbitrary charter duration too often not in line with the WG's
   intrinsics
5. _extremely_ high cost of charter renewal

I was then rather puzzled to see Jeff's summary on current work on
Supergroups does not really seem to address any of these bullet
points. I think that after two years, we need to speed up a bit here.
As an example, the CSS WG is currently rechartering, again, and
almost nothing has changed. Same cost, same endless questions, same ETA
randomness, same time consumption, same complexity. Woof.

I am then proposing to split Working Groups into two different
categories:

A. Longstanding Working Groups
------------------------------

These are Working Groups that we expect to stay for a verrrrry long
time, like html, CSS, DOM, etc. These WGs never end work, they almost
never change scope, they add/remove deliverables all the time and
they're the ones highly impacted by the rechartering cost. Please note
such Groups already deal with Deliverables in a way that is not entirely
in line with the Process since they usually don't officially amend
Charters to start a new work. We really need to adapt
our Process to our evolving practice or our Process is pointless.
For such Groups, rechartering would be a fast-track process and I really
mean it: only an update of the deliverables section. Without Membership
prior request or formal objection from the AC, the Charter would be
"updated" and not "renewed".
The idea is then to keep the same Charter, just updated, with start
and end dates updated too.

B. Single-Purpose Working Groups
--------------------------------

These are Working Groups expected to work on a well-predefined set
of deliverables and shutdown upon completion. They're not expected
to need an extension, although it could happen from time to time.
Rechartering of such Groups would retain the current expensive process.

--

Furthermore, we still have a major issue with the list of Deliverables.
I read many arguments about the usefulness of a list of ETAs for
deliverables. But let's be very clear: in the cases I know, the
complexity of the specs we're dealing with and the sometimes
counter-productive perfectionnism of our Membership make that list of ETAs
mostly unrealistic if not totally random. The ETA for a spec depends on
too many variables, and variables the Consortium and even the Chairs
have no control upon. One of the main gordian knots is the availability
of a Test Suite, a time-consuming effort that Members don't
always easily dive into. That alone makes ETAs too often unpredictable.

We have too often in the past hid ourselves behind IPR-related
rationales to enforce the presence of such ETAs in our Charters. I
think it's time to acknowledge the fact the release of a Standard
is not always a straight line and that even with the best intentions
in the world, a Standard is not a Software project. In some cases,
giving an ETA is doable and reliable; in others, it's not and there's
almost nothing we can do against it. Please note this is an opinion
other Members also expressed. As an example, ETAs for the Deliverables
section of the new Charter of the CSS WG are, *again*, a problem and
a too complex effort. It's mostly useless, sucks our time, and let's
be serious, nobody ever looks at these dates again.

I think it's highly time for a major - and urgent - revamp here. Don't
misunderstand me: IPR-related arguments are of course perfectly fine,
but if they lead to almost random guesstimates that nobody really cares
about, we are only creating frustration and waste of time/energy. And
I don't think we have time/energy to waste.

Thanks.

[1] https://lists.w3.org/Archives/Member/w3c-ac-forum/2016AprJun/0176.html
[2] https://lists.w3.org/Archives/Public/www-archive/2013Sep/0011.html

</Daniel>

Received on Monday, 20 June 2016 13:43:25 UTC