See also: IRC log, previous 2006-10-16
ACTION: [DONE] Ben announce new URIs for editors' drafts to TF [recorded in http://www.w3.org/2006/10/16-htmltf-minutes.html#action04]
ACTION: [DONE] Elias summarize the reification discussion [recorded in http://www.w3.org/2006/10/16-htmltf-minutes.html#action11]
ACTION: Mark write examples/tests of striping support [recorded in http://www.w3.org/2006/10/16-htmltf-minutes.html#action10] [CONTINUES]
Mark: examples is nearly ready to send out
ACTION: Ben update the issues list [recorded in http://www.w3.org/2006/09/19-htmltf-minutes.html#action08] [CONTINUES]
ACTION: Ben start separate mail threads on remaining discussion topics [recorded in http://www.w3.org/2006/09/19-htmltf-minutes.html#action07] [CONTINUES]
ACTION: Steven to put together sample XHTML2 doc with all mime type, etc.. [recorded in http://www.w3.org/2006/09/19-htmltf-minutes.html#action01] [CONTINUES]
Steven: I work a bit on this each week; you can see changes
<Steven> http://www.w3.org/MarkUp/2006/xhtml2.xml
Steven: at the moment it is intentionally
served as text/xml
... I will make a snapshot as .xhtml so
it is served with the correct MIME type
... Also, we've submitted a WWW2007 RDFa workshop proposal
Ben: reminder; we've targetted 27 Oct as date for code and examples of new syntax
ACTION: Elias start an FAQ [recorded in http://www.w3.org/2006/10/03-htmltf-minutes.html#action10] [IN PROGRESS]
Elias: just started it
<EliasT> ... http://esw.w3.org/topic/RDFa#FAQ
Mark: my impression after the discussion was
that some of the problems were due to reification in a particular context
... for what RDFa needs, the issue seems more clear
... e.g. 'who said X' does seem to be clear
... in the references that Elias cited
... Dan's point is that he wants RDFa to be no more expressive than
RDF/XML
... so that all RDFa can be transformed to RDF/XML
Elias: I don't agree that it is clear how we can identify a triple
Mark: if it's not clear how to do this in RDF/XML then we don't need to do it in RDFa
Elias: by 'identify' I mean know whether it
exists in a specific named graph
... it's not clear how to say 'that triple'
... the RDF spec is clear that reification is about identification, not about
quoting
... but we don't have a clear way to identify
Mark: but if RDF/XML hasn't solved the problem then RDFa can't solve it either
Elias: but we have documents
Mark: we can name some other type of Statement
but we can't name an RDF triple
... if we wanted to, we could say there is an xyz:Statement that has implicit
information
... but we'd still need some magic if there's no way to say that this maps to
an rdf:Statement
Elias: RDF reification does not say that if you have the quad it implies the triple
Mark: in terms of generation from RDF/XML
that's the case
... you're not supposed to assume that the original triple exists
... but that doesn't prevent the other way around
... RDFa could say the triple exists
Ben: we can start with some RDFa and specify what it means in RDF/XML
<EliasT> Note that asserting the reification is not the same as asserting the original statement, and neither implies the other. That is, when someone says that John said something about the weight of a tent, they are not making a statement about the weight of a tent themselves, they are making a statement about something John said. Conversely, when someone describes the weight of a tent, they are not also making a statement about a statement they made (since they may
Ben: but we can say that the dc:creator of the s/p/o quad is responsible for a triple asserted by that quad
Elias: we want to attach provenance to HTML documents, not to graphs
Ben: we know how to do that
... e.g. a LINK or a META on a DIV can say who is the creator of that DIV
Mark: even without reification there is a
question
... on the one hand you're making statements about the DIV
... on the other hand we're reusing it at a higher level to represent a
bnode
Ben: but the bnode is just a local concept
Mark: to be consistent, the only thing you'd be making statements about is the DIV
<EliasT> <div class="foo bar"/>
Mark: similar issue to describing the car or the web page that describes the car
Elias: given a div with two classes, can we say that the two type statements were made by different creators?
Ralph: that seems a corner case
Mark: when we embed metadata in the document rather than have it standalone then we lose the ability to attach provenance to the metadata
Ben: we could add more to reify the class="foo bar" case if we really wanted to
<EliasT> <div>
<EliasT> <meta property="class">foo</meta>
<EliasT> <meta property="class">bar</meta>
<EliasT> </div>
Ben: that would be one way to think about it
but this is still more than RDF/XML gives
... but another way is to simply write the triple two times
Mark: 2 choices; 1. use reification as it is
... we've agreed that you can't connect between the asserted triple and the
quad
... 2. say that the "RDF-ness" is not really in the document; explode the
document, expand every property in the document so that the contents are all
objects, etc.
... then this exploded version is converted back 'down' to RDF/XML
... method 2 makes things quite complicated
Elias: even if we go with method 1, there is
still the question -- adding provenance to a DIV adds statements to multiple
triples
... and that DIV could become the object of some enclosing element
<EliasT> <div rel="foo" rev="bar">
<EliasT> .. reification ...
<EliasT> </div>
Elias: if our granularity is the XML element the reification applies to two different things
Mark: I could just about live with this
... but in the IPTC case, given the statement "Reuters says this article is
about skiing. Reuters is 80% confident in this categorization."
... and separately "BBC says this article is about skiing. BBC is 60%
confident in this categorization."
Ben: but this is an RDF problem too
Elias: if rev="bar" was added later by a different person than rel="foo" ...
Ben: you're talking about multiple authors changing the same piece of HTML
Elias: don't confuse two things; an HTML
element and ...
... the HTML element can create many triples
Ben: I can understand that attaching a link to
a DIV is attaching the link to every triple
... this may not be the best practice, but that _is_ what the HTML says
<EliasT> <link rel="iptc:category" href="[iptc:skiiing]">
<EliasT> <meta property="iptc:confidence" content=".8" />
<EliasT> </link>
Elias: Reuter's 80% confidence is not in the
statement itself but that the category is skiing
... if I add iptc:joke to the categories, the 80% should still only apply to
skiing
Ben: but why do we need to support that?
Elias: I am also worried about the case where the LINK element participates in an enclosing rev property
Ben: we need to look at some tricky examples give this new proposal and make sure they turn out sort-of reasonable
Mark: now I'm having trouble seeing these
examples as reification
... I am reevaluating my earlier comments suggesting that n-ary relations
were not what we wanted
... the category/confidence example can be done with n-ary relatoins
... IPTC used a LINK element to add a confidence
... perhaps a way out of the provenance issue is to handle it at the HTML
level
... use an additional HTML element rather than adding properties [to an
existing one]
... I'm starting to see that an n-ary solution fits more cases than I'd
originally thought
... I will extract more examples from original IPTC documents
ACTION: Mark produce more examples of applicability of n-ary relations from IPTC documents [recorded in http://www.w3.org/2006/10/23-htmltf-minutes.html#action08]
ACTION: Elias to send by email examples of HTML markup that could be problematic with reification [recorded in http://www.w3.org/2006/10/23-htmltf-minutes.html#action09]
<EliasT> http://www.w3.org/TR/swbp-n-aryRelations/#useCase1
<Steven> http://www.w3.org/People/Ivan/