Re: embed RDFa --> embed coolness into Yahoo search results

Ian Hickson wrote:
> Some tests that can be passed to Yahoo!'s testing page:
> 
>    http://hixie.ch/www/tests/adhoc/rdfa

So you have 4 tests.

- #1 contains RDFa attributes but no XHTML namespace, so I guess you
don't want this to generate triples. But Yahoo has already decided that
it will parse RDFa attributes in anything it can parse, including
DataRSS, the extension to RSS they've already proposed.

I understand that you might disagree with this decision, but it's a
decision they chose to make, not a bug.


- #2 Yahoo is not seeing any triples, but our parsers are. I think the
issue is that Yahoo, for this application, is looking for you to link
your games to the top-level page using the relation media:game. So this
test works:

http://ben.adida.net/misc/rdfa-test-2.xhtml

I see three issues:

(a) some of your CURIEs (med:ia/game) are not passing muster at Yahoo,
even though they work well with all of the parsers we've built in the
task force.

(b) some of the resources are being resolved relative to the wrong base,
but I'm not sure what's causing that (W3C validator, or Yahoo).

(c) the prefixes like aaaaaa are not being picked up in the UI, although
they are being picked up in the RDF/XML, so something is off here.

I'll check with Yahoo on all three of these.


- #3, you're not linking each item to the top level page with the
rel="media:game", so Yahoo isn't seeing them. Slight tweak just like #2
above:

http://ben.adida.net/misc/rdfa-test-3.xhtml

Notice how both are picked up.


- #4: you're splitting up the triples into multiple spots, but again the
same problem is that you're not linking your games with rel="media:game"
at the right level. Here's the fix:

http://ben.adida.net/misc/rdfa-test-4.xhtml

Notice again how the triples are picked up in the RDF/XML, but I do
think there is one additional issue:

(d) triples defined in different places, although they are in the
RDF/XML, are not picked up appropriately in the Yahoo UI.


So thanks! We've got 4 actionable bugs:

(a) complex CURIEs should work, including "/" in the suffix.

(b) resolving URLs should be done against the proper BASE.

(c) prefixes other than default should be picked up in the UI, not just
in the RDF/XML

(d) triples defined in multiple places should be picked up in the UI,
not just in the RDF/XML


I'll communicate these to Yahoo and get back to you.

-Ben

Received on Tuesday, 17 March 2009 20:00:37 UTC