[Fwd: GRDDL schedule, in retrospect; life-after-REC thoughts]

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

attached mail follows:

The more I think about the GRDDL WG process vs stuff
like the HTML WG and even the TAG, the more I like it.
It was a small, productive group. I didn't think
we'd do ftf meetings nor use cases; I was right about
ftf meetings but wrong about use cases. It
was a solid, test-driven process with pretty robust
decision records.

You might recall I did a "speed of light" schedule estimate
for GRDDL back in Aug 2005.
http://lists.w3.org/Archives/Member/w3c-semweb-cg/2005Aug/0024.html

And of course, the charter estimates various milestones like
last call, Proposed Rec, and Rec.

I just got out a spreadsheet to look at the actual dates
vs my Aug 2005 estimate and vs the July 2006 charter:

http://www.w3.org/2001/sw/grddl-wg/sched-retrospect
http://www.w3.org/2001/sw/grddl-wg/sched-retrospect.html
http://www.w3.org/2001/sw/grddl-wg/sched-retrospect.gnumeric

The actual time from CFP to 1st telcon was 3.7 weeks,
which was 1.5 times what the charter estimated; that's
a slip of 1.3 weeks. If I had it to do over again,
I think I'd estimate that the same way; the alternative
is to pick the date of the 1st meeting in the activity/charter
proposal; that's worthwhile for groups that start with
a ftf meeting, but it's a _lot_ of work, and a 1.3 week
slip is cheap in comparison.

The ratio of actual/estimated for CFP to LC was 1.7,
for a slip of 13.1 weeks. I'll know better than
to think I can skip use cases next time, even
for a mature design. #issue-http-header-links ate
up some of that, too.

The PR milestone slipped 28.1 weeks in total. That's
right around the 6 months of contingency that
is pretty much the norm for W3C charters (it
goes in as "post-REC", initially, but often gets
used up for other stuff). Some of that
could go down to overly optimistic scheduling
(i.e. planning to skip use cases),
but much of it is explained by stuff that was
really not known at the time of charter writing:
putting test cases on the REC track, editorial
resources for the primer coming and going,
#issue-base-param testing follow-up, late energy
on #issue-faithful-infoset , etc.

I think the integration of EARL into our test reporting
work was an unplanned benefit, meanwhile.

The results of the Jan 2006 call for input are a remarkably
robust indicator of how the WG would go. I highly recommend
that tactic. I think the only big surprise in the WG
size and shape since then was bringing on Harry as chair.
http://lists.w3.org/Archives/Public/semantic-web/2006Jan/0251
http://www.w3.org/2002/09/wbs/1/grddl-cfi/results

My Aug 2005 estimate had more CR exit criteria
than we actually went with:

[[
- CR 1 Dec
    We could skip this, since there are already 3 implementations,
    2 running services, and a test suite. But not all the
    tests have been passed by all the implementations, so we should
    look at why not; plus, I think we should
    - add GRDDL support to the RDF validation service (at least
      by way of a link from the RDF validation page to the existing
      GRDDL service)
    - integrate GRDDL into the HTML validation service at some level;
      at least by way of a link.
    - get GRDDL+SPARQL services hosted by 3 communities
    - get Dublin Core's profile to be grddl-happy
]]

If I weren't chairing the HTML WG, I might feel more obliged
to do more of that first-hand. But as it is, I think I'll
leave it to the CG to figure out the priority of GRDDL
post-REC deployment stuff from here on out.




-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/