Re: RDFa test case #1 missing @profile?

/ Shane McCarron <shane@aptest.com> was heard to say:
[...]
| attributes as triggers for triple generation.  Not to belabor this
| point, but do you really think there are documents in the wild that
| use attributes like the ones we have defined AND scoped values
| (CURIES)?  Without that combination, I don't see any rogue triples
| being generated anyway.

I expect it's only the attribute names that matter, right? The values,
whatever they are, would turn up in the triples.

As I said before, my concern is as much about process and precedent as
it is about the specific case of RDFa. That said, I think the RDFa
case is particularly hard because it will extract triples, semantic
meaning from my documents (and some therefore will claim that I am
obligated to uphold those statements).

Suppose I publish a spec tommorow that asserts the meaning of
attributes named 'oid' in all XHTML markup. Suppose further that I
intend to deploy software to a large hunk of hypertext world that
implements those semantics.

I think you might reasonably push back and say I have no authority to
do that. There may be systems out there that use 'oid', and the
current XHTML spec allows them to do so. I don't own the XHTML
namespace, so I have no business defining semantics for it.

The case against RDFa isn't quite as cut-and-dry because you are the
XHTML WG and so you do own the namespace. On the other hand, if you
expect RDFa semantics to apply to vanilla HTML in no namespace, I
think you've overstepped the boundaries almost as clearly as my
hypothetical example.

This is all even more frustrating because there are clear, obvious
technical approaches that completely avoid any of these issues.
The first two that come to mind are:

- Require some sort of explicit markup to assert support for the RDFa
  semantics. I think the DTD is too tenuous, but anything in the
  markup would be fine, any attribute on the html or head elements,
  for example.

- Put the attributes in your own namespace, so that you can be certain
  no one has legitimately used them for something else.

But these aren't considered tractable because of infrastructure
deficiencies in the field and various forms of author confusion and
laziness about namespaces.

| Would you be open to the converse requirement?  Not speaking for the
| task force, just spitballing, and wording to be massaged, but:
|
| A conforming RDFa processor MUST extract triples from any document
| that uses the DOCTYPE declaration, @version, or @profile values
| defined herein.
|
| Document authors who wish to ensure their triples are extracted SHOULD
| use one of these declaration methods.

Yes, I think that would make me feel a lot better.

                                        Be seeing you,
                                          norm

-- 
Norman Walsh <ndw@nwalsh.com> | How is the world ruled and how do wars
http://nwalsh.com/            | start? Diplomats tell lies to
                              | journalists and then believe what they
                              | read.--Karl Kraus