Reference Format for N3?

RDF/XML has N-Triples as a reference format to test against, but N3
has no such counterpart (except for the complicated --reify output
which really just attempts to express all N3 extensions in regular
RDF). A reference format would be low hanging fruit for N3
formalisation.

Three designs I've been considering:

1) Strip down N3 as much as possible.

This is tricky because of formulae, which can't be reduced any more
than to curly brackets. You'd also need to retain @forAll and @forSome
for custom quantification levels.

2) Make something as N-Triples-like as possible.

Again, the main changes that would need to take place would be to
extend the format to accommodate quantification over some custom
level. For this, we'd need shortnames for formulae, in their own
namespace parallel to the bNode and universals shortname namespaces.

The main point is that instead of having the variables be maintained
as lists attached to a formula, as in the CWM backend, you maintain a
link in the syntax from the variable to the formula. A univar might
look like ?var%formulaname. Not pretty, but N-Triples isnt pretty
either; it's a reference format, as easy to parse as possible.

3) Make something simpler than N-Triples.

It'd be nice to have $something for bNodes rather than _:, and to take
those <> delimiters off from around URIs. This occurs to me because
option 2 would require changing the language anyway, so why not go
even further?

I'm leaning towards option 2. The format would have to handle triples
(in the default context), and quads; N-Quiples? It would also be
natural to have ?var quantified over the unnamed root formula no
matter what subformula it appears in, which is rather different from
how things are done in N3 of course. The basic idea is to make
conformance testing easy, but perhaps it'd be useful as an interchange
format too.

Thoughts? Comments? Worth submitting as a CWM RFE?

Thanks,

-- 
Sean B. Palmer, http://inamidst.com/sbp/

Received on Wednesday, 24 October 2007 15:19:57 UTC