comments on RDFa Test Cases

Michael, Wing,

Great work on the test cases. Here are some comments on each one. I'm
looking at: http://www.w3.org/2006/07/SWD/RDFa/testsuite/testcases/

The biggest issue to address, in my opinion, is "META and LINK" in the
body. If we want to test these, we *have* to produce these tests with
extension .xhtml and with mime type application/xhtml+xml, and then IE
may complain loudly. I would suggest having a bunch of straight
text/html tests *without* META and LINK in the body, and then a few
xhtml+xml ones *with* META and LINK in the body.

I also note some bug in your SPARQL in the latter examples... using .ttl
instead of .htm as the subject. Easy to fix :)

Great work!

-Ben

================

(Note that I don't see #1,9,18)


#2
Let's use standard HTML: put a HEAD in there.

#3 and #4 and #5
use of META and LINK in the BODY... This is going to be problematic. You
should deliver these files using the xhtml+xml mime type, otherwise the
DOM is manipulated by most browsers (and Tidy). We should probably
mention this for all the test cases that use META and LINK in the body.

I would recommend staying away from META and LINK in the BODY for the
first few test cases, leaving those to special test cases. That's a big
part of the latest Primer: not requiring META and LINK in the body.

#6
Ahah, CURIEs. You are jumping right into the more controversial test
cases :) How about one that doesn't use a CURIE first?

#7
Great.

#8
Great.

#10
I don't see the SPARQL for this one, but otherwise good.

#11, #12
LINK in the BODY again. Since there's no rendered content here, you
could make this into a test for the LINK in the HEAD, as per Bob
DuCharme's use case. Also, I'm not sure FOAF folks would like you saying
that one MAILTO knows another MAILTO. A foaf:Person with mailto knowing
another foaf:Person, yes.

#13
Hmmm, I don't like this one. You've got rendered content in the HEAD,
which is not super kosher. Also, you use about="" on the HEAD, which is
not supposed to be necessary. What's the point of this test?

#14, #15
Good, good way to show META in the HEAD and the use of xml:lang.
Is there a bug in the SPARQL with that extra ">"?

#16
Great.

#17, 18, 19, 20
LINK and META in the BODY again.


(numbers skip to 101)

#101
Great.

#102
Nice, good first test of striping.

#103
Great.

#104
SPARQL is wrong, it expects a subject of .ttl instead of .htm

#105
META in the body. Same problem with SPARQL subject of .ttl instead of .htm

#106
Great. SPARQL bug, .ttl instead of .htm

#107
Great. .ttl bug.

#108
Oh, very nice striping test. .ttl bug.

#109
Good, seems repetitive with earlier tests.

#110
Good, a bit repetitive, too, though less so (use of CONTENT)

#111
We didn't really define what HREF on DIV would do (except in XHTML2), so
I'm not sure this is a good test case. What are you trying to test here?

#112
Same ">" in the SPARQL... maybe I don't know SPARQL and this is the
right syntax?

#113
Interesting, technically, the spaces around "Mark Birbeck" should be
part of the object literal here, no?

Received on Tuesday, 6 March 2007 18:00:45 UTC