Re: RDF Design (Complaint) FAQ

On 16 May 2013, at 18:16, Pat Hayes <phayes@ihmc.us> wrote:
>>> == Why can't predicates be blank nodes? ==
>> 
>> No sane use case. (There are use cases for literally as predicates, e.g. microdata.)
> 
> Actually that is about as wrong as it can be. Not only are there sane cases, you can't get a complete inference rule set for RDFS *without* allowing this. See Appendix A of the current Semantics draft. 

Defining the semantics of a language, or implementing the language, are not use cases for a language. I don't mind if people write rule sets over generalised forms of RDF, or support such forms internally in their implementations, as long as the standard doesn't require everybody to support it.

> 
>>> * Because some people have written code that assumes they are not. (good  example?)
>>> 
>>> == Why can't subjects be literals? ==
>> 
>> Changing this now would require changing every syntax, every parser, most APIs, and many other standards and tools built on the RDF data model. Some RDF users would like to see this change, but many simply don't find the advantages gained very compelling, and find that the cost of change outweigh the benefits. Therefore, there's some vocal opposition to this change. The way the W3C process works, there's a bias towards the status quo, so the change is unlikely to happen.
> 
> It has to happen at some point, if only because the use cases are so compelling, especially if one has a bias against using bnodes.

See above. There are plenty of people who don't find the use cases compelling at all.

Some people find the use cases compelling. It does not follow that the change has to happen.

> And I don't believe that it would require changing EVERY parser, etc..

Of course it would.

Richard



> The rejection of literal subjects seems to come mostly from RDF/XML syntax, which is going to be of historic interest only in due course. 
> 
>>> * Because some people have written code that assume they are not. (good example?)
>>> 
>>> * Because it would lead more people to bad modeling practice like  { "London" rdf:type eg:City }
>>> 
>>> == Why do we have so many kinds of collections/containers? ==
>> 
>> Doing this properly would require putting lists right into the data model, moving away a bit from the simple set of triples data model. This would be too ambitious a change, given the narrow charter scope, so the group didn't go there.
>> 
>> Any attempt at encoding order in a set of triples is a hack. Different such hacks were tried over time, none of them make everyone happy.
> 
> You can DESCRIBE order in a set of triples, and we do that already. But descriptions, by their nature, can be partial and incomplete. 
> 
>>> * No one's come up with a design that's enough better than all the others to make people be willing to give up the others
>>> 
>>> == Why wont you give guidance about whether to use reification or datasets? ==
>> 
>> A few vocal users who like reification, so no consensus that datasets are better, so we can't say it in the spec.
> 
> Also, more generally, its not the job of a spec to give guidance. 
> 
>>> == Why are language-tagged strings different from everything else? ==
>> 
>> Historical reasons that are very complicated. Didn't find a way to retrofit a cleaner design without breaking too much.
> 
> Every design involves some kind of exception, or some ugliness (trailing @ in PlainLiteral, eg) or a tweak to an existing standard which was beyond our control. 
> 
>>> * Historically, there was a believe that XML Schema datatypes had to be free from language tagging
>>> 
>>> == Why are there unicode characters disallowed from xsd:string literals? ==
>> 
>> They are disallowed in XML. There was an early decision to base
>> the RDF data types on XML Schema types, so we inherited these restrictions when we said that our strings are XSD strings. From an RDF point of view, the restrictions don't make sense.
> 
> Right.
> 
>> 
>>> * Because strings are intended to be displayed;
> 
> Really?? Who said that? I have never heard that idea expressed before either in this WG or the previous one. Strings are, well, strings. They have no preferred purpose. I find this argument bogus. 
> 
> Pat
> 
> 
> ------------------------------------------------------------
> 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 Thursday, 16 May 2013 18:31:46 UTC