W3C home > Mailing lists > Public > semantic-web@w3.org > June 2010

Re: What is it that's wrong with rdf:List [summary]

From: Pat Hayes <phayes@ihmc.us>
Date: Sat, 19 Jun 2010 15:59:10 -0500
Cc: Dan Brickley <danbri@danbri.org>, Norman Gray <norman@astro.gla.ac.uk>, Semantic Web <semantic-web@w3.org>
Message-Id: <8ABCE745-4CF0-4695-8C43-4E64E7B1552B@ihmc.us>
To: Sandro Hawke <sandro@w3.org>
Not that assigning blame several years after the event is really  
constructive, but I think that the current situation was the result of  
what might be best described as a culture clash. Lists were put into  
RDF because the task of encoding OWL syntax made it very clear that  
the RDF containers notions weren't good enough; and that weakness,  
once identified, was clearly of wider importance. So OWL made it clear  
to the RDF WG that RDF needed LIsts; and the RDF WG knew that OWL  
would use them in the way they eventually did. What we did not  
realize, I think - certainly I did not - was that this would result in  
their being made into lower-caste vocabulary, thereby ruling out their  
use for other purposes. The entire ethic and Tao of the RDF team was  
to avoid any such syntactic restrictions or limitations on RDF, to  
keep it as 'open' as we possibly could. (Just like it is in OWL-Full,  
in fact. One might characterize OWL-Full as the way OWL would have  
been if the RDF WG had designed it, instead of OWL-DL, which is the  
way OWL would have been if the description logicians had designed it.  
Except, of course, they *did* design it.)

Its important to remember, when trying to understand how cock-ups like  
this can occur, that each WG had its own charter and internal  
dynamics, and they didn't always exactly coincide. To many in the OWL  
WG, the chief point was to define OWL as a description logic, and the  
RDF encoding was at best a necessary nuisance and at worst a positive  
disaster. (PFPS is *still* arguing his passionate anti-RDF case, using  
the same specious arguments about how it produces paradoxes.) To the  
RDF WG, the point was to build some early version of TBL's 'layer  
cake' with RDF at the bottom and RDFS and OWL sitting on top of it. If  
one of these layers happens to be a description logic, more or less,  
then great; but that wasn't the primary point. These two points of  
view often clashed, sometimes with considerable energy, and the result  
was an obvious compromise. And one can see how the difference in  
perspective on how Lists would be treated can have arisen. From the  
RDF POV, letting OWL use LIsts would just be in the general spirit of  
let a thousand flowers bloom. From the DL perspective, the key issue  
is to maintain the internal purity and cohesiveness of this particular  
logic, and maintaining a sharp distinction between logical and non- 
logical vocabulary, protected in software with syntax checks and error  
messages and so forth, is centrally important. Just using RDF to  
encode the syntax was already, for many in the group, like asking them  
to swim in a sewer. But to allow people to do arbitrary things with  
the very syntax in which their logic was written was like asking them  
bathe naked. We in the RDF WG were already bottom feeders, so we  
didn't even see what their problem was.

Pat


On Jun 19, 2010, at 11:16 AM, Sandro Hawke wrote:

> On Sat, 2010-06-19 at 17:05 +0200, Dan Brickley wrote:
>>
>>
>>
>> On 19 Jun 2010, at 16:47, Norman Gray <norman@astro.gla.ac.uk> wrote:
>>
>>>
>>> Greetings.
>>>
>>> Many thanks, everyone, for interesting and thoughtful explanations
>>> here.
>>>
>>> In summary, it seems that:
>>>
>>> * rdf:List isn't fundamentally defective, but
>>> * it is often at least inconvenient to use with SPARQL, and
>>> * it collides with OWL, which has hijacked rdf:List for internal  
>>> use.
>>
>> We added rdf:List to rdf in the 2004 rdfcore revision, to address  
>> some
>> owl-related needs. So 'hijack' is a little unfair :)
>
> I think the negativity and frustration of 'hijack' is valid, although
> the implications aren't quite right.  rdf:List may have been motivated
> in part for use by OWL 1, but as I recall it was put in the RDF
> namespace because it was intended for general use.  I'm trying to  
> think
> of some replacement for hijack that suggests "shot itself in the  
> foot".
>
> How about:
>
>   * it collides with OWL DL, which uses rdf:Lists internally but  
> cannot
> be used when other rdf:Lists are in use.
>
> I think that puts the onus where it belongs, on OWL DL, instead of on
> rdf:Lists.
>
>     -- Sandro
>
>
>
>

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
Received on Saturday, 19 June 2010 21:01:10 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:36 GMT