Re: Minutes: Syntax subgroup teleconf - Oct 26, 2001

> I'm sorry to see an important decision like this being made at a
> telecon where not everyone was present.

My understanding is that the Syntax subgroup did not make a decision for the
WG, but decided to recommend a course of action to the WG. Since considering
this recommendation was not on the agenda either last week or this week,
your response is not too late.

However, ...

As the one abstention on this at the syntax subgroup telecon, I feel that a
strong case needs to be made to go against the majority opinion there.

I feel the basis of that opinion was essentially that my approach was
overkill. Almost all specs, in some way or other, depend on structured
English, and typically they suffice. Just because M&S was in poorly
structured English we shouldn't regard the technique as tarnished.

Afterwards, despite my disappointment, I felt that in order for the WG to be
different (i.e. significantly more formal than most specs.) we need to
(collectively) have enthusiasm for the project of greater formalism than
structured English. We do not. There has been little engagement with my
attempts to formalise the triple production better.

There still is one possibility, that M&S was not only poorly structure
English, but poorly structured ideas. This might make it well nigh
impossible to write clear structured english that accurately describes
anything that approximates to the M&S triple production rules. bagID and
aboutEach are the key candidates to watch. If, with all care and attention,
we cannot articulate clearly what these might mean then we should consider
either dropping them, or articulating them with my formal approach (which I
believe can give a meaning to them consistent with M&S).

I think another problem with M&S was the lack of an effective reference
implementation. SiRPAC was just too buggy, and its active endorsement by W3C
added to the confusion. I am firmly opposed to confusing reference
implementations and specs. Both are useful, but they serve different
purposes, and should be distinct. The need for an 'executable' spec is an
attempt to have the best of both worlds, which typically is unnecessary.

Thanks for your support Aaron,

Jeremy


PS Bill is mistaken in thinking that my approach requires XSLT, technically
it doesn't. Certainly it doesn't require understanding the pages of
autogenerated XSLT.

Received on Friday, 2 November 2001 09:31:57 UTC