Re: ISSUE-1: Status of RDFa Profiles

On 2010-3-12 24:39 , Ben Adida wrote:
> On 3/11/10 9:01 AM, Manu Sporny wrote:
>> * The profile document can specify both tokens and prefixes
> 
> I'm still worried about inheriting prefixes from @profile, so I'm not
> sure we have consensus just yet. I think we still have an issue
> resolving both backwards compatibility and copy-and-paste'ability if we
> allow for profiles to define prefixes. I don't think Mark and Ivan's
> responses have resolved this.
> 
> Let me restate the issue I'm worried about: Google offers a wizard that
> lets you create a rich snippet chunk that you paste into your page, with
> @profile importing a "product" prefix.
> 
>  <div profile="http://rdf.google.com/rich-snippets">
>    <span property="product:name">Canon Digital Rebel XTi</span>
>    <span property="product:type">Camera</span>
>    <span property="product:price">$500</span>
>  </div>
> 
> Now you paste that into your page, but you forget that somewhere higher
> up in the DOM, you declared a conflicting @xmlns:
> 
>  <div xmlns:product="http://rdf.yahoo.com/product#">
> 
>     ... stuff you paste ...
> 
>  </div>
> 
> Given this situation, we have the following choices:
> 
> (1) @profile overrides parent @xmlns ==> RDFa 1.0 generates triples that
> are different from those defined by RDFa 1.1.
> 
> (2) @profile never overrides parent @xmlns ==> @profile is incompatible
> with copy-and-paste.
> 
> Given that we agree that @profile can appear anywhere, I don't see
> another option, and both suck.
> 
> Mark's solution of keeping different prefix lists generated by @xmlns
> and @profile is equivalent to solution (2), which kills copy-and-paste
> when we precisely want users of @profile to not have to worry about
> complexity and @xmlns.

I agree. I would certainly not want option (2)

> 
> Ivan's point that we have the same situation with keywords is
> interesting, but I think not quite right, because
> 
> (a) the number of default keywords is small and predictable, unlike the
> number of possible @xmlns prefixes declared in a random user's page
> 

I accept that, though...

> (b) web sites preparing copy-and-pasteable chunks of HTML+RDFa can thus
> take care not to override existing HTML4/5 keywords.
> 

... I myself do not remember all the HTML4 keywords:-( I am not sure it
is a perfectly safe assumption to simply rely on web authors here.

But again, I think my main argument was that you are trying to optimize
against an RDFa1.0 processor processing an invalid RDFa1.0 file. The
result of your cut-and-paste

<div xmlns:product="http://rdf.yahoo.com/product#">
  <div profile="http://rdf.google.com/rich-snippets">
    <span property="product:name">Canon Digital Rebel XTi</span>
    <span property="product:type">Camera</span>
    <span property="product:price">$500</span>
  </div>
</div>

is an invalid RDFa1.0 file, indeed invalid XHTML+RDFa1.0. If the author
does that _and_ uses RDFa1.0 processor then indeed it may lead to
confusion but do we really want that?

We could reinforce the usage of @version and requiring RDFa1.1
processors to ignore @profile altogether if the version is 1.0...

> 
> I understand that, if @profile does not include prefixes, then we're not
> providing the convenience function that Ivan wants for RDF-community
> authors. But I think the potential confusion / loss of copy-and-paste is
> too high a price to pay for this convenience.

I am still not convinced. And the inconvenience of large portions of
xmlns statements (I have over 15 in the header of my foaf file), with
the associated risk of misspelling is a major burden that I really would
like to reduce.

> 
>> * Are we limiting next/prev/index/license/etc to @rel/@rev or allowing
>>    them everywhere?
> 
> I think that, in RDFa 1.0, they are limited to just @rel and @rev,
> right? 

correct

>        So that's actually nice because it means that the copy-and-paste
> keyword confusion between RDFa 1.0 and RDFa 1.1 is limited to just
> @rel/@rev, which makes it, once again, an even smaller problem.
> 
> I think it's okay for us to say that new @profile's can define keywords
> for all attributes.
> 

I can live with that, absolutely.

>> * What is the mental model are tokens/prefixes two different concepts
>>    in RDFa or are they the same thing?
> 
> I don't think we'll be able to convince the average user that they are
> the same thing. I appreciate the elegance of Mark's proposal, but that
> elegance is at a level of abstraction that I think most HTML authors are
> not familiar with.
> 

I agree

Ivan


> -Ben
> 

-- 

Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF   : http://www.ivan-herman.net/foaf.rdf
vCard  : http://www.ivan-herman.net/HermanIvan.vcf