Re: why S doesn't require double properties [was: Datatyping Summary V4]

On 2002-02-05 1:47, "ext Pat Hayes" <phayes@ai.uwf.edu> wrote:


>> People must be free to use either idiom, and since folks
>> *will* be using both idioms, even if not in the same
>> knowledge base, there will *have* to be two versions of
>> every ontology, one for each idiom.
> 
> Well, it may not be this bad. Here's a worst-case scenario. There are
> ontologies out there using all the idioms, and you want to be able to
> get information from them all and interrelate it all together. One
> way you can do this is to choose the idiom you like, and translate
> the others into that one on the fly. For local typing this is easy;
> you could do it with a simple scripting language. (You have to be
> able to recognize urirefs denoting datatypes, but those are supposed
> to be 'global' in any case, right?)
> For range-style typing you have an option,  but the easiest way to
> handle it (and the safest) would be to  try to infer a 'local'
> statement of type from the range information and then, as a
> precaution, applying your transformations rules on the local
> assertion (that you have managed to conclude) to make sure it is in
> the local idiom you prefer.
> 
> If everyone is prepared to do this much work, then you can publish
> your RDF using any idiom and be sure that people will understand you.

I am not willing to do that much work if it is not necessary
to do so. And I believe that the KISS principle very much
applies here. If we *have* to use different vocabularies for
different idioms, so be it, but I do not concede that we must.

One of the attractions (for me at least, and I think for others
as well) of RDF is the ability to freely syndicate knowledge
in order to achieve a synergetic knowledge base. Having to do
any gymnastics on syndication to prevent datatyping idioms
from colliding is IMO unacceptable if it can be avoided.

Users deserve our every effort to avoid such complexity.

I will go down kicking and screaming on this issue. Sorry.


>>>>  Issue B5: Storage Requirements
>>>>  ===============================
>>>> 
>>>>  status: disputed.
>>>> 
>>>>  TDL requires significantly more storage to implement.
>>> 
>>>  Sergey got back on this one, no?
>>> 
>>>  In short: you may not need to store the whole string
>>>  lots of times, but you do need to store some sort
>>>  of distinct identity for each occurence of a string.
>> 
>> We've already clarified that there are means to allow
>> TDL to support tidy literals, both by idiom as well
>> as MT. I believe that this is a non-issue, or at most
>> an issue nearing resolution.
>> 
> 
> I agree. The only case that absolutely requires untidy literal nodes
> is the old P++ case, where an in-line literal is used 'directly', but
> what it denotes is sensitive to property ranges, as in
> 
> <mary> <age> "10" .
> <age> <rdfs:range> <xsd:integer> .
> 
> meaning Mary is 10 (not "10") years old. Apart from this, we can have
> tidy literal nodes.
> 
> Pat

Can we close this issue? Sergey?

Patrick


--
               
Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email: patrick.stickler@nokia.com

Received on Tuesday, 5 February 2002 04:13:20 UTC