Re: Datayped tagged literals: a case for option 4 vs option 2d

On 28 Sep 2011, at 12:00, Steve Harris wrote:
>> The nice things about extensibility points is that the misdesigned and bizarre extensions just get ignored.
> I used to agree with this opinion, but now strongly disagree.
> The problem is when misdesigned/bizarre extensions get used in some areas, so can't be deprecated, but still cause endless confusion for people coming to RDF anew.
> Good examples are RDF containers (used in RSS) and RDF collections (used in OWL). They're both pretty nasty to work with, but can't realistically be removed, and appear at first glance to solve a real problem.

I didn't mean extensions that are baked right into the core spec – what I said above doesn't apply to them, as you rightly point out. I meant third-party extensions.

>> (Aside: As an extension mechanism, datatypes in RDF are a miserable failure. Outside of the core RDF and OWL specs, *not a single* new datatype has been defined that has actually caught on and become popular.)
> Not sure I agree, XSD already provides a wide variety of useful types, and it's rare that people have to step outside that. It is true that non-XSD types are pretty rare in the wild, but I'm not sure that qualifies as a failure.

RDF datatypes provide an extension point. Almost no one uses that extension point. Those who do, use it either in inappropriate ways (DBpedia), or could easily use a different mechanism (e.g., intermediate blank node, or bake the syntax into the property definition). Thus, RDF could have been designed without that extension point (just hardcoding the XSD types, similar to JSON) and would have been simpler without sacrificing anything.

Of course we know this only with hindsight, and it's too late to change now, so whatever…


Received on Wednesday, 28 September 2011 11:05:28 UTC