Re: New version of Atom+RDFa

On Mon, 21 May 2012 16:59:28 +0200
Ivan Herman <ivan@w3.org> wrote:

> - Process comment: personally, I would like to see this document end
> up as a W3C WG Note.

As would I.

> 7th of June. This means moving the document to W3C space, ask for a
> short name from Thomas and... that is about it...

That's rather short notice, but I'm already using ReSpec.js, so it's
probably not a lot of work.

> 1. The document defines a default vocabulary URI. It is, however,
> silent on whether the RDFa Core initial context terms, defined in
> http://www.w3.org/2011/rdfa-context/rdfa-1.1 are valid for Atom as
> well or not. There is no inheritance of those terms, so it must be
> stated explicitly. I would think these should be valid for Atom, too.

The RFC 4287 explicitly says that when (for example) rel="alternate"
occurs in an Atom feed, it is just a shorthand for
rel="http://www.iana.org/assignments/relation/alternate". Atom was
ahead of RDFa in putting compacted URIs into @rel by several years. :-)

RFC 4287 says that this is the case for anything that matches
"isegment-nz-nc" as defined by IRI (effectively the same as terms in
RDFa). Thus rel="next", rel="previous" and so on are also expanded to
URIs in the same namespace.

It's my intention that this remains the case under Atom+RDFa 1.1, so
the default XHTML+RDFa terms cannot have effect. It's my understanding
that @vocab "wins" over the default terms now, however it probably
needs to be made clear that (in the case of Atom), the host
language's default vocab also "wins".

I would however want Atom+RDFa to inherit the namespaces from RDFa Core.

> 2. I am not sure I understand the rationale for the rules in the
> <link> element. Can you explain?

There are basically two restrictions on link[@rel] in Atom+RDFa 1.1.
You can't use CURIEs, and you can't use terms in the scope of a
default vocabulary except http://www.iana.org/assignments/relation/.

To put it concisely, it's because in either of those situations, an
Atom+RDFa 1.1 processor would expand the CURIE or term to a different
URI than a plain Atom 1.0 processor would.

For example, rel="foaf:homepage" would be expanded to
<http://xmlns.com/foaf/0.1/homepage> by an RDFa processor, but to
<foaf:homepage> (it's a valid IRI syntactically, but not a registered
scheme) by Atom processors.

Similarly, rel="homepage" with vocab="http://xmlns.com/foaf/0.1/" would
be expanded to <http://xmlns.com/foaf/0.1/homepage> by an RDFa
processor, but to <http://www.iana.org/assignments/relation/homepage>
by Atom processors which would ignore @vocab.

There may well be other restrictions that make sense to add: for
instance, forbidding @about on <link>.

> I should have my Atom implementation updated soon in pyRdfa...

Mine still follows the old spec, but I intend to update it tomorrow.

-- 
Toby A Inkster
<mailto:mail@tobyinkster.co.uk>
<http://tobyinkster.co.uk>

Received on Monday, 21 May 2012 16:31:53 UTC