Re: RDF 2 Wishlist

On 2009-11-02, Pat Hayes wrote:

> That was my first reaction, but on reflection this really is a good 
> idea. ALL literals should have a datatype. Having plain literals in 
> RDF is a very clear example of premature optimization. So +1 to Jie.

In here I'd say that, yes, it wouldn't break much if a plain literal 
would get an implicit type after the introduction of literal types. 
Simply make it a default and introduce a formal logical equivalence. I 
sincerely doubt anybody would've produced code which would fail after 
this sort of conservative modification. Especially since typed literals 
are such a hack in the first place, and as such often neglected in the 
first place...

I really hate the idea of literals to start with, and I've said so in 
the past. From the get go, they didn't have much of a semantic flavour, 
while at the same time we already had standards in place to handle 
literal and other kinds of un-/semistructured data. They had much more 
semantics defined than string literals ever did in RDF. If we really, 
really wanted to include them into the RDF data model, we could have 
done it using inline MIME references from the RDF triple store to 
included literal data obeying external semantics. But now that we have 
what we have...

Within the confines of the triple model, I'd like the extra convention 
that typed literals can and *must* always be represented as blank nodes 
with an untyped value hanging off a related triple, and an extra 
assertion telling the type of the base, untyped literal. Either the idea 
of a typed literal should be deprecated, or it should be made strictly 
equivalent to this kind of regular representation.

The same also goes for n-ary relations; I don't think the current 
specification for that sort of thing is good enough, because it permits 
multiple different kinds of representation, and doesn't present 
axiomatic semantics which would enable unification into a single normal 
form. Which then makes it rather cumbersome to handle n-ary relations in 
a systematic fashion in code. Please, make the presentation schema where 
there is a blank and/or named node at the top, and multiple attributes 
(named by predicate labels) hanging off it the standard representation 
of a relational tuple.

Let's then call those two things my primary wishes for RDF2. ;)
-- 
Sampo Syreeni, aka decoy - decoy@iki.fi, http://decoy.iki.fi/front
+358-50-5756111, 025E D175 ABE5 027C 9494 EEB0 E090 8BA9 0509 85C2

Received on Monday, 2 November 2009 20:43:30 UTC