[Xesam] Validator

[Antoni Mylka]

Ivan Frade pisze:
> Hi Antoni
> 
> On Mon, Jul 27, 2009 at 5:39 PM, Antoni Mylka <antoni.mylka at gmail.com
> <mailto:antoni.mylka at gmail.com>> wrote:
> 
>     Ivan Frade pisze:
>     > 1) Release a tarball with the ontologies + HTML documentation so
>     > distributions can build packages from it, and tracker, strigi,
>     > kde-nepomuk depend on that package.
>     >     For this we should use python/C and autotools. We have already a
>     > "soft validator" in C, plus a basic HTML generation also in C.
>     >
>     > This is the point we discuss on GCDS, and agree on the svn layout, use
>     > autotools and write the HTML generation.
> 
>     No problem as far as I'm concerned. Two questions though
>     - what was the agreed way to add examples/testcases to the 'ontologies'
>     folder.
> 
> 
>  No, the idea is just to release the ontology files in a TRIG format in
> a well-known directory ($(PREFIX)/ontologies IIRC) + some HTML doc (so
> you can browse the ontology docs offline)
> 
>  No examples or test-cases for the ontology. There is no code: just the
> ontology definition.

I meant examples <=> testcases. Each example is a test case that is
either valid or invalid. If we change an ontology in a way that one of
the example becomes invalid - we've broken the test case, now the
example will have to be updated to reflect the change in ontology. It's
as simple as it gets. If we have 100 examples that were valid in version
x, and all of them are valid in version (x+1) without changes, then we
can say that (x+1) is backward-compatible with (x) because any
application that expects the data to be similar to those examples won't
break.

> 
>     - do you intend to generate header files / vocabulary constants /
>     whatever is the c equivalent to a class similar to
>     http://tinyurl.com/nxfe5c, or leave that up to application developers?
> 
> 
>  No, the package only provide the ontologies. Each project decides what
> to do with them: tracker uses the ontologies to generate the DB schema,
> other projects have a generic triplet store and use the ontology in
> their own way... The same applies to libraries: it is not clear how to
> expose SparQL in a programatic way, so each library can decide how to
> generate its own API.

OK.

>     >  I couldn't check the wiki page, but i guess antoni wrote down what is
>     > his validator checking so i can cross-check what is missing in the C
>     > version.
> 
>     Is there something wrong with the trac wiki?
> 
> 
>  For some reason i cannot access to the links published in the mailing
> list (it asks for password). I can access going to the project web page
> and accessing trac from there.

Drop the 's' from 'https' at the beginning. BTW it seems you're not
added as the project member on sourceforge. What is your SF username, so
that I could add you? Then you will be able to access the https urls too.

>     >  What is blocking me now to write a proper HTML generator is to decide
>     > what language to use. I can complete the C code, or rewrite the things
>     > in python (nicer HTML template engines, easier string handling...).
>     > Python looks nicer but i am not sure there is a turtle parser...
> 
>     By all means I'm for python/ruby (actually ruby more, but python will
>     not be much of a problem either). All of them have redland bindings, and
>     raptor does claim to support turtle.
> 
> 
>  I never used ruby but i feel comfortable in python. I need to check the
> redland bindings.

I'd say, hack together some initial version and commit it to the oscaf
subversion, say tools/doc-generator then we can all tweak it.

Antoni Myla
antoni.mylka at gmail.com