Mikael,

I wasn't talking so much about the problem of how to build instance  
data as I was talking about how to build an rdfs/owl ontology using  
property ranges that support 'correct' inferencing and, by extension,  
xml schemas to enforce it. And I wasn't trying to suggest that  
explicitly defining additional subproperties was a good solution. I  
was actually trying to suggest that it was possible to define dual or  
null ranges in a base property that would _allow_ the declaration of  
more explicit subproperties in application profiles. Please forgive  
me for not being clearer, but I was unsure of my audience and didn't  
want to get into an extended discussion of the dcam and dcaps. And I  
should have referenced the post[1] of Karen Coyle's that got me  
thinking about it in the first place in order to provide a bit more  
context.

--Jon

[1] http://kcoyle.blogspot.com/2008/01/more-on-rda-and-literals.html

On Mar 17, 2008, at 3:55 PM, Mikael Nilsson wrote:

> mån 2008-03-17 klockan 15:26 -0400 skrev Diane I. Hillmann:
>> Folks:
>>
>> To enhance your reading pleasure this week, as you're looking at the
>> "First pass at RDA elements" I already sent, I'd point you to my
>> colleague Jon Phipps' blog post of yesterday:
>> http://jonphipps.wordpress.com/2008/03/16/simple-dc-and-rda/
>>
>> This post grew out of conversations we've been having concerning the
>> necessity to make it possible for those transforming legacy data  
>> (with
>> textual values) to be able to use RDA, without compromising the  
>> ability
>> for those using URIs.  This has to do with the ability to  
>> accomplish all
>> the JSC Scenarios using RDA.
>
> Thanks for the pointer.
>
> The proposal makes technical sense. As he notes, this would be  
> somewhat
> of a mirror of how DCMI ended up handling their situation (though  
> with a
> very different historical origin).
>
> There are two issues that I can see immediately.
>
> 1. Creating two or even three versions of each property would be a  
> hard
> sell, politically. And confusing, to users of the properties.
>
> 2. I'm concerned that the proposal might be aiming to solve a problem
> that isn't really a problem. I'm assuming from your and Jon's
> descriptions, that the problem is people who are in the following
> situation:
>
> "I have this literal value, and this RDA property with a non-literal
> range. What do I do?"
>
> And, the DCAM gives you the answer. You create a statement with the  
> RDA
> property, using NO value URI, no VES, and a single value string. In
> DC-TEXT [1]:
>
> DescriptionSet (
>   Description (
>     Statement (
>       PropertyURI ( rda:publisher )
>       ValueString ( "Acme Publishing" )
>     )
>   )
> )
>
> I fail to see what problem is *not* solve by the above.
>
> Thus, my answer would be NO, we don't want to do that. Noone else  
> in the
> RDF community needs to, and neither should we. DCMI had to for legacy
> reasons, but we'd only create interoperability issues if we went down
> the "multiple properties meaning the same thing" route.
>
> /Mikael
>
> [1] http://dublincore.org/documents/dc-text/
>
>
>
>>
>> Please feel free to comment on the blog and/or on these lists.
>>
>> Thanks,
>> Diane
>>
>> *********************************
>> Diane I. Hillmann
>> Director of Metadata Initiatives
>> Information Institute of Syracuse
>> Email: [log in to unmask]
>> Voice: (607) 387-9207
>> Fax: (607) 793-9505
>> Skype: dihillmann
>> *********************************
>>
> -- 
> <[log in to unmask]>
>
> Plus ça change, plus c'est la même chose