Re: LC Issue timbl-01 choices

>>>Dan Connolly said:
> On Tue, 2003-04-01 at 07:12, Dave Beckett wrote:
> > timbl-01 is about the first question Tim raised in
> >   http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0226.html
> > about the use of bagID, we had some discussion of it last week but
> > didn't reach a conclusion or discuss in the last telcon.  
> > 
> > 
> > The first question in the email is:
> >   "Is this feature then worth implementing? What does the group think?"
> > 
> > We know bagID it isn't used, and probably could be killed if we
> > didn't feel such a change was rather late and/or constrained by
> > charter.
> 
> The charter constraint didn't seem to bother us that
> much in the case of aboutEach and such. (though
> there was some action to check with the SemWeb CG
> whether it was OK, and I'm not sure what, if anything,
> ever became of that.)

The charter constraint did bother us and we did consider it
significant, hence the CG action at the same time when we made the
decision in
   http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Dec/0089.html

We cautiously made the change and reported to the community in the WD
of December 2001:
  http://www.w3.org/TR/2001/WD-rdf-syntax-grammar-20011218/
with very little negative feedback since.  We got 1 LC comment about it
  http://www.w3.org/2001/sw/RDFCore/20030123-issues/#chas-01


bagID has been implemented from these WDs and test cases.  We have the
evidence from the number of people that implemented the parser test
case entailments.


> 
> >   If we wanted however that would mean new last call
> > documents for RDF/XML Syntax and Primer - removing things.
> 
> How do you come to that conclusion?

It is a substantial change - more than editorial and I thought it was
polite to make all the substantial changes before CR.  We would also
need to write an explanation why this change would be made.

It would be good to get the community to buy into such a change so
that we can leave CR stage.  At present, we have several complete and
correct implementations of the rdf/xml to rdf graph mappings so could
match probable CR exit requirements right now (for syntax WD at
least).  If we change the mapping, we suddenly don't have any
implementations.


> The only way to find out for sure that you need another
> last call is to ask to proceed to the next step and
> get turned down.

Which hat is Tim wearing here?  Let's not waste our time :)


> I suggest that removing it in the course of moving
> to CR/PR would be welcomed by reviewers and implementors.
> 
> There's an argument to be made that a last call is in
> order so that anybody who's using it has a chance
> to say "hey! I wanted that! put it back!" But it's
> not a black-and-white issue. If, as you say, we
> know bagID isn't used, then there's little reason
> to do a last call.

s/do a/do another/

> 
> >   In this
> > case the answer would be:
> > 
> >   No and we accept your comment. We will remove bagID from the
> >   language and WDs.
> 
> I think you can stop there.
> 
> >   This substantial change will mean preparing a
> >   new set of last call documents.
> 
> That question seems separable.

I'll defer to you on the detail of
  http://www.w3.org/Consortium/Process-20010719/tr.html#last-call


> Or perhaps you're saying that it's your considered
> opinion that the WG should do another last call if
> it removes bagID?

I really care nothing about this bit of syntax, I've never used it.

My concerns are the WDs that will need changing syntax and primer,
existing implementations and documents that mention it.  For our WDs,
I suspect I might have the easier job than the RDF Primer editors.
The implementation change should be ok, removing code.  There is at
least one book being published right now that will be documenting
RDF-as-revised including rdf:bagID.   I know these are are all
"work in progress" but it is a concern.

> > 
> > The alternative is to reject this.  I don't see the point of
> > postponing this any further, we didn't remove it, wrote test cases
> > for it and people implemented it.
> 
> Yes, the implementors understand and were convinced that it's
> worth implementing.
> 
> I think TimBL has raised an interesting question as to whether
> the users understand what they're saying when they use this
> construct.

How do you find that out?  Another WD that asks specifically?

<snip/>

Dave

Received on Tuesday, 1 April 2003 11:13:09 UTC