Re: RDF Recommendation Set comments (re agenda for 6th April)

fensel wrote:
> - I do not think that the problem of RDF is that it does not provide enough.
> For a base layer were other languages are layered on top it provides actually
> quite a lot of features. Bags, collections, and lists, reification, and statements
> in the language over other language statements etc. Frankly, I think a better
> job would have to keep RDF as simple and basic as possible. So I would
> rather love to see a simplification than an enrichment of RDF.

Personally, I'd like to see a simplification of RDF (like dropping three 
of the things you mentioned), and an enrichment of RDFS (not ness. 
including sameAs though).

I agree with Pat, that presenting a single set of specs which is enough 
for people to cover most use cases with (without saying go here for OWL 
X,Y,Z but note these difference between x,y,z and our own s, there for 
literals, elsewhere for datatypes, this special one for this flavour of 
the month spec at time y) would be beneficial. Especially around core 
properties people use daily for basic tasks and creating ontologies/schemas.

> - Why include stuff in the very base layer of the semantic web that you know
> that other languages will forbid to keep inference tractable.

Does that relate to all of the properties that could be folded in to 
RDFS? (see http://www.w3.org/2009/12/rdf-ws/papers/ws31) or just that 
sameAs one? or a few?

> - Proper equality on the web is a fascinating research question and I expect
> more and more PhDs on it. Currently its complexity is still mostly neglected or
> "solved" very simplistic as in OWL. Again, I wonder whether a research issue
> should already be standardized at the most basic level. Many people use
> owl:sameAs because there is nothing better around for their purpose (similar SKOS
> stuff comes without reasoning support). I do not think they are really happy
> with it and calling it rdf:sameAs will not solve their problem either.

sameAs is a real PITA, especially when you take in to account that most 
people actually want to say "in the triples you get from my graph <here> 
I'm using <name1> to talk about the same thing <namefoo> in the <there>" 
and people want to consider it in the same quady who-said-it way and do 
things like "oh yes, I trust <there> so I'll accept the statements they 
say about <namefoo> as being about <name1> too, apart from this one 
which makes no sense". Rather than the jammed down your through data 
munging approach that says "<namefoo> and <here> are names for the same 
thing, go forth and merge those statements now!".

Other properties aren't quite so (..insert word here..), for instance 
differentFrom pretty much just says what everybody thinks anyways given 
two different URIs, equivalentClass/Property are somewhat milder (even 
in name) and so forth.

sameAs, is def somewhat of a special case, and could be quite dangerous 
when dealing with data from different sources on the wild web, the 
others, not so (imho).

Best,

Nathan

Received on Thursday, 7 April 2011 19:04:45 UTC