Re: [RDFa] ISSUE-28: following your nose to the RDFa specification

On Thu, 2007-06-21 at 12:12 -0700, Ben Adida wrote:
> Dan Connolly wrote:
> > This issue is about the case where you do find an RDFa
> > attribute (about/resource/etc); how do
> > you know that the author meant it in the RDFa sense?
> 
> [...]
> 
> > No, at worse you get some triples out that the author didn't intend
> > because s/he didn't mean the markup to have RDFa meaning.
> 
> You have to be realistic: separate the spec from what people do in
> practice. Take microformats for example. Their spec says that users must
> have a PROFILE attribute. Deployments show that this is almost never the
> case, meaning that microformat parsers out there are assuming tons of
> stuff about what authors say when they didn't "intend to."
> 
> I don't see you arguing that these parsers should stop what they're
> doing, rather I see you arguing that people should add profiles.
> Similarly, we're not going to tell people to stop looking for RDFa, but
> we are going to tell RDFa publishers to put the right DTD declaration
> and we are going to tell the HTML WG to make their spec extensible
> enough so that publishers can do the right thing and declare their usage
> of RDFa.

Ah.. so negotiation with the HTML WG _is_ in the critical path for RDFa.
I've been struggling to get a clear answer about that (and
other requirements questions).
The schedule you recently gave only goes up to Working Draft.
I suppose this negotiation should happen before Last Call; at the
vary latest, it has to happen before Proposed Recommendation.
http://www.w3.org/2006/07/SWD/wiki/RDFa

Perhaps a release of RDFa scoped only to modular XHTML could
go to REC first, and then an update could integrate
with HTML 5.

I wonder what sort of "done" you had in mind when
making this estimate...

 Ben: we should aim to be done before November
  -- http://www.w3.org/2007/05/31-rdfa-minutes.html#item03



> (It is interesting to note that if the HTML WG chooses to not make HTML
> extensible, then that will clearly encourage ad-hoc HTML extensions
> where you *can't* follow your nose.

Let's hope that doesn't happen...

>  You can't stop the desire for
> semantic markup, you can only attempt to guide it.)
> 
> If you're arguing that we should tell people to strictly *not* look for
> RDFa unless there is a DTD or profile, then I strongly disagree.

I'm arguing that the spec should say that if a document doesn't
have a follow-your-nose connection to the RDFa spec, then
consumers who infer RDFa meaning from the markup do so at their
own risk.

In many cases, such as a users clicking an RDFa bookmarklet,
that risk is completely reasonable. The user can tell the
machine "hey... that looks like RDFa; go ahead and treat
it like RDFa; if it wasn't meant that way, I'll accept
the consequences."

> In an ideal world, folks will move to a DTD/schema where the RDFa syntax
> means something explicitly every time. If they don't, assuming RDFa
> markup is certainly no worse than assuming 20 different microformat
> profiles.

Right; the risks are pretty much the same.

>  In fact, I consider it a much lesser offense, since we've
> already modified the RDFa syntax a few times so that it is extremely
> unlikely for someone to write RDFa syntax without actually intending to.
> It is a lot easier to accidentally use a "reserved" microformat class name.
> 
> As to the copy-and paste issue... there's a point that you seem to miss
> when you conflate DTD/mime-type and GRDDL PROFILE. With GRDDL and
> microformats, the PROFILE is *necessary* to figure out what the author
> means: you can copy and paste ONLY if you modify the HEAD of the
> document with each copy-paste operation to keep the *meaning* around. If
> all pages are RDFa enabled with the right DTD, then copy-and-paste works
> just fine: the HEAD of the document does not need to contain any
> vocabulary-specific information, so it does not need to change with
> every copy-and-paste operation. In other words, the isse is not whether
> you have to modify the HEAD of the document, it's whether that
> modification is vocabulary-specific. In the case of RDFa, it's not. That
> makes the inevitable promiscuous parsing situation a bit more reasonable.

Yes, RDFa is more uniform and the situation with GRDDL and
profiles is more complex.

If I understand you correctly, that's a "no" answer to
my question: Is the DTD optional? And likewise for the MIME type.

So copy and paste works as long as all the places people want to
paste to use the new DTD and MIME type. I suppose that's a
reasonable answer as long as RDFa is based on modular XHTML.

I said I'd like clarity on requirements such as "RDFa markup
must work when pasted in the middle of an HTML-as-she-are-spoke
document". I take it that's a goal for a future version of RDFa,
but the current version is only intended to meet the requirement
"RDFa markup must work when pasted in the middle of
an XHTML 1.1+RDFa document with the application/xhtml+xml MIME type."

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Friday, 29 June 2007 23:20:10 UTC