See also: IRC log, previous: http://www.w3.org/2009/07/30-rdfa-minutes.html
<scribe> ACTION: Ben to summarize architectural issues (security) re: @profile proposal for tokenizing the web. [recorded in http://www.w3.org/2009/07/30-rdfa-minutes.html#action04] [DONE]
<scribe> ACTION: Shane to produce proposed diff re: XMLLiteral change [recorded in http://www.w3.org/2009/07/30-rdfa-minutes.html#action03] [CONTINUES]
<scribe> ACTION: Ben to author wiki page with charter template for RDFa IG. Manu to provide support where needed. [recorded in http://www.w3.org/2009/05/28-rdfa-minutes.html#action10] -- continues
benadida: have made some progress on that, got up early to work on it.
<benadida> worth reviewing --> http://rdfa.info/wiki/Rdfa-ig-charter
benadida: might want to have
people start reviewing that now.
... Focuses around two activities.
... Provide input to HTML WG on RDFa.
... RDFa in XHTML v1.1
... The proposals we're making for simpler syntax (@profile,
@token) will push us towards 1.1 for XHTML+RDFa
... Those are the two foci of the RDFa IG.
http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2009Aug/0037.html
benadida: That's the summary of the issues...
markbirbeck: @profile is how do we find a bunch of tokens.
benadida: We didn't go too deep
without you here, Mark.
... The security issue seems to have struck a chord with Manu
and Shane.
... The parser now needs to dereference 3rd party resources and
that is an issue with Javascript and XMLHttpRequest.
markbirbeck: Right.
... The substance of my proposal was about the tokenisation
aspect.
... We should consider using the tokenization term.
... We should attempt to address the xmlns: issues that people
have been concerned about.
... The issues that you are raising are with the mechanism of
finding other tokens.
benadida: The question is where in the stack does the tokenization happen?
markbirbeck: I think tokenization
does exist now.
... This is purely about languages that reflect RDF.
... You're saying that tokenization is part of the CURIE
stack.
... Language (N3, RDF/XML) has its own way of expressing
RDF.
... CURIEs do a pretty good job of summarizing URIs.
... Why not allow the syntax to not need colons for expressing
CURIEs.
benadida: Right, so to be
precise: In any implementation, where would it live?
... This would have to live in the layer above the RDF
store...
... You have to resolve the CURIE into a URI before you put it
into the RDF store.
Shane: Mark's proposal is in
essence, in an existing RDFa document today, with a minor
change to the processing model, you could declare a token
mapping /in-line/ and then use that token mapping as a defined
term.
... you could do something like <span token="shane=http://example.org/shane#"
property="shane">Shane McCarron</span>
benadida: I think the common goal we want is the simpler syntax.
<ShaneM> <span xmlns:shane="http://www.halindrome.com/me" rel="shane">Shane McCarron</span>
Manu: These are separate issues, tokenization and @profile.
benadida: Well, what's the goal?
markbirbeck: So, there are a
number of goals.
... One of them is that xmlns: is problematic.
... We're trying to make it easier to author.
... The big problem with Microformats was never really the
syntax.
... The vocabularies don't scale.
... There isn't a unified parsing mechanism.
... The goal of tokenization is to provide Microformats-like
markup.
... I'm proposing a solution for what we agree is a particular
problem.
... Tokenization isn't an end in itself.
... it isn't.
<Zakim> ShaneM, you wanted to talk about why remote vocabs are a problem
ShaneM: I don't think anybody
disagrees that we need to provide an easier markup
mechanism.
... The issue isn't @token - it's that having collections of
those terms expressed remotely in a way that is discoverable,
is problematic
... it's problematic because we have issues with Javascript +
XMLHttpRequest.
markbirbeck: Yes, I understand
that.
... There are ways of loading remote URLs in Javascript.
... We still haven't addressed the conceptual question...
... There are conceptual issues about mapping at the RDF
level.
... We need to get the conceptual level right.
benadida: I think we agree on the
approach.
... I think it's a symptom of the conceptual issue.
... I think we're putting too much responsibility on the
parser.
... The parser has to dereference 3rd party resources.
... It's a symptom of the larger issue.
markbirbeck: I fundamentally
disagree with you.
... There is an issue - how do you get the document?
... But that doesn't tell you anything about the fundamental
distinction.
... We have a way of mapping text to URIs.
... currently in RDFa.
... So, using owl:sameAs is problematic.
... All @token does is add a way of mapping text to URIs.
<markbirbeck> Slightly modify Shane's example, by adding a colon:
<markbirbeck> <span xmlns:shane="http://www.halindrome.com/me" rel="shane:">Shane McCarron</span>
markbirbeck: So, we can already
to the markup above.
... When you're working out CURIEs, the above works.
... The proposal doesn't hinge on @profile or @token - it's
just a way to use colon-less reserved words in
@property/@rel/etc.
benadida: I don't think we can
look at that goal separately from making the syntax
simpler.
... If we're looking at the end-goal - looking at the proposal
piece-meal doesn't help.
markbirbeck: What I want to stress is that the rdf:subproperty approach requires some mechanism to load an external document.
<benadida> google's markup:
<benadida> <div xmlns:v="http://rdf.data-vocabulary.org/#" typeof="v:Review">
<benadida> <p><strong><span property="v:itemReviewed">Blast 'Em Up</span>
<benadida> Review</strong></p>
<benadida> <p>by <span property="v:reviewer">Bob Smith</span></p>
<benadida> <p><span property="v:dtReviewed">March 20, 2009</span></p>
<benadida> <p><span property="v:description">This is a great game. I
<benadida> enjoyed it from the opening battle to the final showdown
<benadida> with the evil aliens.</span></p>
<benadida> <p><span property="v:rating">4.5</span> out of 5 stars</p>
<benadida> </div>
<benadida> <div prefix="http://rdf.data-vocabulary.org/#" typeof="Review">
<benadida> <p><strong><span property="itemReviewed">Blast 'Em Up</span>
<benadida> Review</strong></p>
<benadida> ...
benadida: You have to dereference
at some point.
... But the key is, /when/ you do it.
... Mark, in your proposal, until you dereference the 3rd party
resource, you've got nothing that you can do any processing
on.
markbirbeck: If you went the
@profile route, then yes, you are correct.
... There are different ways of doing it.
... One way is where @profile is hardwired or well-known.
... Your proposal achieves something different than where I see
the whole Microformats-style going.
<ShaneM> Note that these two approaches (token and default prefix) are complementary.
markbirbeck: provided that they don't use any of the standard terms (such as next/prev/etc). Example: prefix: reference
markbirbeck: With your proposal,
its very difficult to mix vocabularies.
... doing this is difficult if not impossible: profile="a=b
c=d"
... with your proposal.
benadida: There is some way that
we need to resolve these reserved words.
... To fully resolve it, I'd rather we delegate it to the RDF
stack.
... Rather than pushing that functionality into the RDFa
processor.
... Adding another layer of resolution in the parser is going
to cause us more problems than solve issues.
markbirbeck: That is a separate
discussion - how do we load other stuff.
... Even if you forget the owl:sameAs stuff.
... One issue is how we map things.
... how do we map yahoo's products to google's products.
... I'm not talking about that.
... I'm talking about simply the abbreviation mechanism.
... my criticism of your proposal, is that there are 2 major
criticisms.
... one - you're using suffixes instead of tokens
... we came up with an algorithm in XHTML+RDFa so that you can
distinguish those from values.
... problem two - you only solve the problem for one vocabulary
in any given context.
benadida: problem one is a small
weakness - I agree.
... problem two, I disagree with.
... Google can, in their schema, say v:title maps to
dc:title
markbirbeck: What happens when I
want to add properties from the Good Relations
vocabulary?
... how can I mix vocabularies?
benadida: That's not a problem,
it's a feature
... We're addressing authors that want to just use one
vocabulary.
markbirbeck: We're getting to the
crux of the problem we're trying to solve.
... Most may not understand how to mix vocabularies.
... Most of the work I'm doing requires that you use many, many
vocabularies.
... We should work out what we're trying to achieve.
<Zakim> ShaneM, you wanted to disagree about what curies know about references and to mention that authors can create hybrid vocabs too
ShaneM: I disagree that CURIEs
don't know about references.
... we built into a CURIEs a way to specify a default
prefix.
... reserved words are also processed first.
... I think being able to set the default prefix can be
enabled.
benadida: What are the next steps?
markbirbeck: There may be a way of harmonizing the two.
ShaneM: The two proposals are complimentary.
<ShaneM> I wonder if it is possible to define a hybrid "vocabulary" that encompasses all of the existing microformat terms?
markbirbeck: Ben, your proposal
is that we enable default prefix.
... If we're happy that @token is resolved, then we may be able
to enable default prefix.
benadida: Worried about teaching these features.
markbirbeck: I think we focus on the language elements that address the use cases.
<ShaneM> reserved terms are checked first in the RDFa processing model... If we permitted "tokens" then those would get evaluated before looking for a default prefix. Ben seems to agree with this.
benadida: I'm worried about the consequences to the stack with your solution... what role are we giving the parser.