RE: Summary and some analysis: New Semantics Initiative

> From: Brian McBride [mailto:bwm@hplb.hpl.hp.com]
> At 14:01 12/06/2002 -0700, R.V.Guha wrote:
> 
> [...]
> 
> >  I am not 
> surprised at the 
> > reaction of Peter F Patel Schnieder and Ian Horrocks. They 
> have in the 
> > past made it very clear that they do not, as a matter of 
> principle, like 
> > axiomatic definitions of languages.
> 
> Fair enough, and I should make it clear that Peter was very 
> open that his 
> motivation derived from a stylistic concern that axiomatic 
> representations 
> of semantics are more error prone.

I think the RDF Core wg needs to be responsive to the collective
opinon of the WebOnt WG (as well as ourselves and others). I'm
not comfortable making a decision about this if its a matter of
personal tastes.

Yes, we should factor 'risk' into our decision, but unless they
can prove something is actually wrong with the solution the risk
does not mean we must kill the proposal.


> >  What we are proposing is relatively old-hat. There may be 
> discussions of 
> > style, but not of substance. So, I don't expect much time 
> to get spent on 
> > this.

I've made this mistake before. 

As a process matter, if we are going to add something important
late in the game, we are also going to allow adequate time for
reviewing it. Its either important enough to face the unpleasant
need to extend the schedule or it is not. I will oppose any effort
to get both.


> I've got a sense that people in the WG are supportive of 
> this

That is also my sense. (Speaking personally, I am supportive, but
I'm also scared of a group extending its own scope and scared
of adding something big late in the day. So I fall back on
process as a guide. We need to make a good faith effort to address
the concerns of both the pro and con sides.)

> >  even if it does not get used by webont, I humbly propose 
> that we do.

If this is proposed as solving a need expressed by WebOnt, then I think
we are on solid process grounds for extending our schedule. If it
is not something they say they have to have, then the case for adding
this gets much harder.


> I want to take the sense of the WG on this, but I would like 
> the WG to know 
> that I have procedural concerns:
> 
>    o this is being rushed - and the need for speed is not 
> clear to me.  I 
> believe that many of the issues of M&S arose from changes 
> made late in the 
> process

Massive agreement.


>    o it is being put together by only a subset of the 
> community interested 
> in this area which does not provide an adequate basis for consensus.

Agreed. My constituents couldn't care less about a debate on
axiomatic definitions of languages. If we have a clear call from a
lot of people saying "there is a problem" then we need to look at it.
If we have a clear statement from experts that "solution B seems
to solve the problem" then we should adopt it. 50/50 arguments
mean this has to drop.

However, please note that the reason we have worked on a model theory
is to catch errors we would not catch otherwise. If we are being told
that we are catching such an error, then we do need to give particular
credence to that.


>      I suppose I might as well be open about one of my 
> worries here.  If a 
> subset of the logicians go one way in RDFCore, and a 
> different subset go a 
> different way in webont then we will be left with a war 
> between the WG's.

Sounds about right.

>      I still think that the right way to do this is to lock all the 
> logicians in a room without a lavatory with all the beer they 
> can drink 
> until they agree on a solution and advise both WG's accordingly.

Tempting. But to be tiresomely pedantic, a more practical way is to
set a deadline and say we can wait until then. (We also need some
guarantee against a minority deliberatly stalling things so it would
run over deadline).

> 
>    o the non-normative status of the proposed new document means that 
> webont can't build a normative spec on it and that does not seem good 
> enough to resolve the issue that is the motivation for doing this.

If its not normative and core to RDF, this group does not have to do it.

> I would like to hear the WG's thoughts on:
> 
>    o to address this problem the solution must be normative; 
> W3C notes and 
> non-normative appendices don't hack it.

as above, i agree

>    o the consensus process must be open to all with a stake 
> in the issue

Define 'stake'. I want a consensus input from WebOnt on the problem(s)
they think we must solve. I'd like a consensus input from them on whether
particular proposals solve those problems. I want third parties
to have a chance to participate. I want our group to have input on this
as well (see model theory point above).

We collect all the input and make the decision. The rest of the world
still has review and comment periods if they don't like the decision.


Ron

Received on Thursday, 13 June 2002 13:51:41 UTC