W3C home > Mailing lists > Public > w3c-rdfcore-wg@w3.org > September 2003

Re: Fwd: Re: comments on 26 September version of RDF Semantics document

From: pat hayes <phayes@ihmc.us>
Date: Tue, 30 Sep 2003 11:01:08 -0500
Message-Id: <p06001f4abb9f51d748f9@[10.0.100.25]>
To: Dan Connolly <connolly@w3.org>
Cc: w3c-rdfcore-wg@w3.org

>On Tue, 2003-09-30 at 05:00, Graham Klyne wrote:
>>  At 20:00 29/09/03 -0500, pat hayes wrote:
>>  >I tend to agree with Peter about the rules being normative, particularly
>>  >as I couldn't find a formal record of that decision either (the best I
>>  >could do was to link to the IRC log). [...]
>
>My memory says we decided to make them normative, but my memory
>isn't very reliable.

That is my memory also.  We talked about it but I think we got so 
deep into the discussions about whether or not to keep reflexivity 
and so on that it got lost in the woods. However on reflection I now 
think we should have considered that step (the normative bit, not the 
reflexive) more carefully, as I don't think it really makes sense to 
say that any set of rules is normative. I think at the time I was in 
an "I'm just the editor, tell me what to do, Im tired"  frame of mind.

>
>The minutes corresponding to the 27 June IRC log are
>http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2003Jul/0025.html
>and they don't seem to show a decision to make the rules normative.
>
>Sigh.
>
>[...]
>>  This makes me wonder if, given that there is less implementation experience
>>  of inference based on these formal semantics, it wouldn't be more
>>  appropriate to request the formal semantics go to CR (with informative
>>  rules) rather than PR at this time?
>
>I think we've been sufficiently careful to add tests for all the
>interesting nooks and crannies in the semantics that passing
>all the tests is quite a bit of implementation experience with
>the formal semantics as written.
>
>Moreover, I maintain that horn rules is a correct implementation
>strategy, and I consider anything to the contrary a bug.

?? Like, translating into FOL and using resolution is a bug? Or using 
a tableaux reasoner (like the industrial-strength Manchester OWL 
reasoners) is a bug? Or using a subgraph-cluster checker for handling 
very large sets is a bug?  (etc.)  That seems silly.

I think of the horn rules as a kind of reasoner-erector-set. You can 
put it together fast, you have the parts pre-built, you don't need a 
shop, and it even works pretty well; but its not necessarily what a 
professional would use.  (Which is fine, of course, provided we don't 
make professionalism illegal.)

Pat



-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes
Received on Tuesday, 30 September 2003 12:01:10 EDT

This archive was generated by hypermail pre-2.1.9 : Tuesday, 30 September 2003 12:01:15 EDT