Re: pyRdfa and the RDFa Test Suite

Hi Ed,

(Michael, I explicitly copy you because there are issues/questions
regarding the test cases that should be settled; thanks to Ed for
bringing these to the fore!)

This is a useful addition! I have stored your setup files and others on
my local disc. My problem for now is that the changes in pyRdfa are
still way too frequent to make it so full packaged with proper versions.
I would like to do that when both the syntax document and the
implementation becomes a bit more stable and I will certainly used that
setup at that point! Thanks!!!!

Of course, you touched on my own vanity:-) and I looked at the tests
that did not work, running them manually one-by-one. _My_ overall
judgement is, however, that most of the problems are related to the
SPARQL (I may be wrong, of course!). Indeed, except for a few cases
(noted below) my claim is that the pyRDFa is fine and conform to the
test; in my view, the SPARQL processor should have accepted it.

There may be two problems/reasons:

- a genuine SPARQL error on our tests. Did we ever checked those with a
processor, in fact?

- problems with RDFlib's SPARQL parser or implementation. A few words on
that one: the irony is that the 'core' SPARQL processing in rdflib is my
making... But then, lack of time, I never made a parser for the SPARQL
language itself. Others did that, but because they used a binary parser,
I could never run that on my windows machine... But the latest SPARQL
test results (that the DAWG is preparing of the Proposed Recommendation
phase of SPARQL) reveals that there are some more genuine problems with
it...

I would probably prefer to run the results on a more reliable SPARQL
implementation, eg, sparqler... Maybe my simple interface to SPARQL[1]
may help in that.

Another comment: the test cases include a number of tests which are
either not yet approved or, worse, rejected and/or on hold. Note to
Michael: it might be a good idea to remove these from the manifest at
some point...

Thanks again Ed,

Ivan

[1]
http://ivanherman.wordpress.com/2007/07/06/sparql-endpoint-interface-to-python/

P.S. Just for completeness, my remarks below. All others either pass or,
again in my view, have SPARQL problems...

> --
> 
> test_0004 (test.XhtmlTests) ... FAIL

Test is on hold but, in fact, to be removed/rejected. We decided not to
use @xml:base

> test_0005 (test.XhtmlTests) ... FAIL

Test is rejected

> test_0011 (test.XhtmlTests) ... FAIL

There is a pending question (in my mind at least) whether XMLLiterals
should be "canonicalized" or not. This is the reason of the failure: I
copy the XML Literal verbatim, whereas the test canonicalizes it.

> test_0017 (test.XhtmlTests) ... FAIL

Although this test is marked as 'on hold', it looks reasonable to me.
But, indeed, it fails on pyRdfa, I will have to look into that...

> test_0022 (test.XhtmlTests) ... FAIL

This is unreviewed, but is definitely to be rejected. We decided not to
use @id.

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf