Re: RDFa 1.1 Issues raised by Google

On Mon, 24 Oct 2011 00:37:45 -0400
Manu Sporny <msporny@digitalbazaar.com> wrote:

> RDFa Lite proposal:
> http://lists.w3.org/Archives/Public/public-vocabs/2011Oct/0093.html

Specifying RDFa Lite as per above as an authoring subset seems fine,
but specifying it as a subset for consumers (i.e. you can support RDFa
Lite by just implementing these attributes and these features) would
be, as far as I'm concerned, a http://enwp.org/Wrecking_amendment .

This is because you'd end up with certain constructs that would mean
significantly different things depending on whether it was parsed as
RDFa 1.1 or RDFa Lite.

Consider this (admittedly obscure) snippet:

	<p typeof="foaf:Person">
	  <span property="foaf:name">Alice</span>
	  knows
	  <span rev="foaf:knows">
	    <span property="foaf:name">Bob</span>
	  </span>
	</p>

Under RDFa 1.1, this means that something called "Bob" knows a person
called Alice.

A theoretical RDFa 1.1 Lite consumer, if it completely ignored the
presence of @rev, would read it as saying that there exists a person
who goes by the names of "Alice" and "Bob".

OK, so maybe you don't like @rev and don't want to honour the
foaf:knows triple there, but ignoring the fundamental fact that Alice
and Bob are different people is pretty bad. This kind of inconsistency
between consuming agents would threaten the viability of RDFa as a
platform to publishing data.

As I say, as a subset which authors can choose to follow or not, RDFa
Lite sounds fine. But if you're targeting this ideas at consumers, it's
potentially very harmful.

Some time ago, I wrote a wiki page on which subsets of RDFa it's safe
to consume. This needs a little update to take into account the changes
in RDFa 1.1 drafts since last year, but should still be reasonably
useful.

http://www.w3.org/2010/02/rdfa/wiki/Subsets

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

Received on Monday, 24 October 2011 16:33:19 UTC