Re: Proposed changes to address issue-dbooth-3 (ambiguity)

On Tue, 2007-06-19 at 13:27 -0400, Murray Maloney wrote:
> At 11:46 AM 6/19/2007 -0400, Harry Halpin wrote:
> 
> >Before proposing any alternative solutions, I would write the Note and
> >specify *all* (including non-normative) spec changes your first proposal
> >recommends before writing a second proposal.
> >
> >The option of reducing variability by *not permitting* Xinclude
> >resolution by the transformation language or the GRDDL-aware agent would
> >require re-opening our decision on #faithful-infoset, which the current
> >proposal in my opinion does not require. At this late stage, I am
> >against re-opening previous WG decisions unless there is very strong
> >feeling by more than one member of the WG, and preferably WG consensus.
> 
> There is not consensus on the suggested change. I am strongly opposed.
> This proposal would require re-opening Faithful Infoset.

I also see an interaction between the proposal and the WG decisions
on issue-faithful-infoset, but in discussions with Harry and Dave,
I didn't find a compelling argument. Reasonable people might
disagree about whether discussion of the proposal should
happen only after re-opening issue-faithful-infoset.

In any case, the testcases Jeremy has put together show
that proposed design is somewhat novel w.r.t. others we have
looked at, and there is some support for it from WG members
other than David.

So I support the agenda Harry has proposed, which includes
discussion of the proposal.
http://www.w3.org/2001/sw/grddl-wg/weekly-agenda 1.151

And I hope we can keep discussions of process to a minimum.

> The current specification is correct to operate on an XPath node tree.
> XSLT, which is our assumed default transformation environment, employs
> a node tree. In all of our discussions of the logical consequences of our 
> decisions,
> we have assumed an XPath node tree which implies XML Normalization, including
> whitespace handling, internal entity resolution, namespace-qualification,
> character encoding, and so on. David's suggested changes would entail 
> changing a large
> number of variables which have not been discussed or accounted for.

Jeremy said something similar...
"I think in general, that late design changes are usually bad design
changes."

I thought the change was analagous to a design I had proposed
a long time ago; then Jeremy's test cases showed a wrinkle
I hadn't considered. That sort of thing makes me wonder how
much QA this design change needs.

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

Received on Tuesday, 19 June 2007 20:31:35 UTC