Re: RDF 2.0 proposal: contextual properties

Hi Jeremy!

Jeremy Carroll wrote:

> Michael Schneider wrote:
>> 
>> Dan Brickley wrote:
>> 
>>> I dislike the existing reification vocabulary. But we can't rewrite 
>>> history: those classes and properties are in use. Should we really make
>>> http://www.w3.org/1999/02/22-rdf-syntax-ns#predicate and the others be 
>>> 404? or at least not describe those terms in the main RDF namespace?
>>>
>>> I'd be happy to see them depractated using OWL or other terminology
>> 
>> LOL! You might be interested in the fact, that RDF Reification has just 
>> been introduced into the current OWL-1.1 draft for mapping the new 
>> concept of annotated (i.e. commented) axioms to RDF syntax:
> 
> I do  not think this was a good idea.
> 
> The semantics of RDF reification is sufficiently broken that this move 
> does not do what it is intended to do, except for the member submission 
> docs not having an OWL Full component, so that they do not pretend to 
> have an RDF compatible semantics. Given that the (draft) charter 
> emphasises the need for such compatibility I am assuming that what to do 
> about axiom annotations will be a moderately difficult issue for the new 
> group.
> 
> I think deprecating reification would usefully mark it as not fit for 
> purpose. It would be helpful if such deprecation was aligned with named 
> graph standardization, which offers a non-broken replacement.

So, let's suppose that in the end, RDF2 is out, reification has been deprecated, and named graphs have made their way into the new version of the standard. So what have we won?

In the past, I heard many people making a lot of strong arguments against reification. At the top of the list was the argument that RDF reification was simply:

    A) useless (or, at least, these people were unclear about what it was good for).

This argument came then in conjunction with other arguments:

    B) The actual semantics in the spec are broken,
    C) reification is restricted to single RDF triples,
    D) uses a lot of space ("triple bloat"),
    E) performs slow, at least if not specially handled in triple stores,
    F) has an ugly and bloaty synax, especially when used with SPARQL,
    G) everyone (including the boss) hates it.

But now with RDF2, we have named graphs, which attacks most of these problems:

    B') have well established semantics,
    C') support arbitrary graphs,
    D') can be stored very compactly,
    E') allow rapid processing,
    F') have a nice syntax, which integrates especially well with SPARQL,
    G') everyone and his dog really love them for killing reification.

Well, this really looks like progress to me: After all, 6 of 7 of the above nagging issues have finally been fixed by replacing RDF reification with named graphs... :-)

Bye,
Michael

-- 
Michael Schneider <m_schnei@gmx.de>

Received on Tuesday, 14 August 2007 07:49:24 UTC