Ben, the difference is that while GRDDL makes it possible to have a
rel value mean something non-obvious (i.e., based upon information
outside of @rel and @profile), RDFa requires it.
On 05/03/2009, at 10:32 AM, Ben Adida wrote:
Maciej Stachowiak wrote:
Microformat-defined rel and class values have their usual semantics
regardless of whether one links a GRDDL transform converting them
to RDF.
How is mnot going to figure out what those semantics are to generate a
proper link-type header? Will all microformats be added to the IETF
link-type registry?
I've read over this thread a few times, and I still haven't seen any
technical argument against the way RDFa handles @rel that is
consistent
with specs prior to RDFa. We have an example with GRDDL (and also with
eRDF, though it's not a w3c spec) that @profile may define an
*indirect*
way, using other elements and attributes, to interpret @rel. RDFa is
no
different.
Julian argues that GRDDL is not about interpreting @rel, it's just
about
extracting RDF/XML. I don't see the difference, but if one wants to
draw
a line, then simply put RDFa on the GRDDL side and assume that it's
"just a way to extract RDF/XML." I think you'd be missing out on how
much you can get out of RDFa, but certainly if GRDDL gets a pass on
this, then RDFa should, too.
In fact, remember that RDFa also specifies @about so you can, for
example, have multiple images each with its own unique copyright
license. For link-type to do the right thing, it actually needs to
fully
parse the RDFa. I'd be excited to have the link-type spec do that,
but I
doubt that's within its scope. So maybe ignoring RDFa is the right
approach for link-type.
-Ben
--
Mark Nottingham http://www.mnot.net/
