RDF-in-XHTML TF -- 17 Jan 2006

Ben: if DanBri shows we have another agenda item to discuss -- the schema for html rel values
... I've make all the critical changes to rdfa-syntax
... the rules are consistent with everything we've discussed
... issues remaining may be around the examples

new examples from Mark

-> Additional Examples for RDF/A Primer [Mark]

Mark: my goal was to start from the viewpoint of someone who is already an HTML author and wants to 'beef up' his document
... starting with blogs, as FOAF works well with them
... got diverted on discussion with DanBri on distinction between document and person and which URI applies to each
... I've come to a new radical conclusion that I will write up
... no longer basing RDF/A on the URI of the document
... a common use case is: given an item of some type, I want to add properties to them
... e.g. an item for sale
... for a FOAF home page, the URI shouldn't identify the person
... could have two FOAF pages at different URIs talking about the same person
... so the actual URI of where the document is is not such important information
... even the information about a license is in another document

[Ralph thinks Mark's realization is plausible; will have to see what implications he draws from it.]

Ben: perhaps a better example for bloggers is that of having a blog be its own RSS feed
... people will understand this
... I'm willing to write up this example for consideration
... I am convinced that more examples makes the document better; thanks, Mark, for pushing this

Pat Hayes' comments

point 1, on the use of "content"

Steven: "content" comes directly from HTML4
... it's always been there and it shows the actual relationship between the element content and the thing we're referring to
... its meaning has not changed from HTML4
... it's useful for people to see that these are natural uses of HTML

Ralph: "content" is a natural word for folks familiar with HTML and that is our target audience

ACTION: Ben respond to Pat's RDF/A comments [recorded in http://www.w3.org/2006/01/17-swbp-minutes#action01]

Steven: re point 2, we can say that yes, we agree; IRIs are the new URIs but we don't need to rename
... in XHTML2 we kept the name URI even though we're changing the value range to IRI
... we say "thing called URI in XHTML2 _is_ an IRI"

Hayes point 3 -- on including "pure" RDF triples

Steven: I could not see how Pat came to the conclusion that <li> was required

Mark: some of our examples use LI -- perhaps those are the only ones that produce pure triples

Ben: let's give another example [to Pat] that shows how to produce a triple without LI
... perhaps another example on "pure RDF" could be added to the Primer
... may be worth reviving the "advanced topics" section of the Primer

Mark: second part of Pat's point relates to Gary Ng's comment on use of html:about rather than rdf:about
... Pat is suggesting that in-line RDF/XML might be useful
... I think it's appropriate to respond to Pat that yes, sprinkling LINK and META throughout the document is the way to introduce RDF triples
... could even link to some embedded RDF

Ben: I'd have gone for a separate document in such a case
... if the author's goal is not to use the rendered content for dual-purpose of metadata, then put the RDF in a separate document
... so respond to Pat with examples other than LI and remind him that <link rel='meta' ...> still works

so RESOLVED

Hayes point 4 -- using CSS to attach properties

Ben: propose to respond that yes, we agree it would be nice for CSS to become RDF/A aware but that is out of our scope

Mark: I think he's suggesting using CSS selectors to attach properties
... our response could be "why not just do that in RDF"?
... e.g. it's not CSS's job to attach an 'age' property to every span

Ralph: CSS's job is all about presentation and RDF/A is _not_ about presentation
... I think extending the use of CSS for semantics is muddying the water

Mark: and this complicates Jeremy's implementation; he'd have to parse the CSS too

Ben: I don't believe we resolved an old issue as to whether class and role attributes affect RDF semantics

next XHTML draft

Ben: is there anything the HTML WG needs from this Task Force at this time?

Steven: no, I'm happy with the state of the document right now
... we're working hard on a new XHTML2 draft
... hope to be able to announce a new draft worth looking at by the end of this week
... we're working on the whole draft, not just the RDF/A bits
... we discovered a number of issues that were recorded as resolve in the issues list but we hadn't replied to the authors

Ben: Bjorn's email from a year ago has been on my to-do list and I think we can now give a more precise response

(ACTION? Ben)

Gary Ng's comments

Re: [ALL] RDF/A Primer for review [Gary Ng 2006-01-16]

section 2, "RDF/A itself"

Ng 2.1 -- on synchronization of metadata

Steven: this is not a new problem and it will never go away
... e.g. the HTML might claim an image is GIF but it can later change its representation to JPEG

Ralph: this touches on the general SemWeb question of 'trust' -- which metadata do you believe? that question is out of scope for XHTML

Ben: in our response we can note that the same problem exists in RDF/XML

Steven: and it exists even in pure HTML; the HTTP headers can disagree with the HTML content

Ng 2.2 - consistency

Ben: answer is to use separate files

Mark: Gary's suggestion here is to reuse the RDF attributes directly
... the reason this fails is because RDF/XML has a striped syntax
... so you don't know whether you're parsing a resource or properties about that resource

Steven: so the answer is that these syntaxes are different

Mark: a lot of work went into trying to resuse the RDF attributes but this failed due to the striping issue

Ng 2.3 - inheriting via nesting

Ben: seems to be a misunderstanding; there is inheritance if there is no new 'about' attribute

Steven: idea is to nest META elements, which you can do

Ng 2.4 - img element

Steven: response is that XHTML2 does allow content in IMG now
... but we didn't resolve the question whether the triples apply to the src

<benadida> TF issues list

Ben: we have a tentative answer to issue 6 of 'yes', pending a further evaluation of the semantics of 'src' from Mark
... given that XHTML2 says the content is ignored unless the src fails

Steven: in terms of processing XHTML2 says the content should be ignored if src succeeds
... given nested IMG with a JPEG src inside an IMG with a PNG src ...

Ben: can we distinguish between parsing and resolving?

Steven: could create some difficulties, depending on whether the nested elements have src attributes
... could continue to process the children for semantic content
... but this would lead to two ways to use the content
... this mixes up the two bits of processing in a messy way

Ben: this same thing happens for objects

Mark: could specify that META and LINK are the first things to be recognized
... even before deciding whether the IMG succeeds or fails

Ralph: I prefer the perspective that says the semantics are still intended even if the content is not used for "rendering"
... for example, a client that is configured not to display images should still consider the metadata to be asserted

Mark: does this answer Gary's question?
... a processor that understands RDF/A can make use of the nested value

Ben: but note that the content is _not_ rendered if the src succeeds, so an author cannot expect the metadata to always be rendered

Mark: if both src and about are specified, we generate both sets of triples
... by the general rule that we generate everything possible

Ben: so advice to authors might be not to specify both src and about [unless you're sure that's what you intended]
... does the HTML WG agree that parsing the content of IMG for metadata is OK even if the src succeeds?

Steven: yes, this isn't a problem

Ben: so we can say that issue 6 is fully resolved as 'yes'?

[no objections]

Steven: I'm pretty sure it's OK but will think about it

ACTION: Steven confirm an answer on issue 6 [recorded in http://www.w3.org/2006/01/17-swbp-minutes#action02]

Mark: appears to be the middle choice

<danbri> http://www.w3.org/2005/05/hrel/ dangling in limbo... can we get a couple reviewers from the tf?

Ng 2.5 - flexible referrals

Mark: doesn't feel true to me
... we've been working on XForms for a while

Ben: maybe another example would help

Steven: what Gary is asking for in forms is already provided in XHTML forms by using P3P
... P3P is an annotation vocabulary specifically targetted at privacy
... Gary's use case has already been covered

Ben: seems we could just use meta and link to annotate elements
... I don't see that there is a need for special subjects and objects

Mark: could imagine some XLink-like language
... but it gets messy to add a level of indirection to every element

Steven: Gary seems to be asking to be able to talk about attributes in the same way we can talk about elements

Ben: it's a question of complexity
... can annotate forms elements

Mark: Gary may be trying to find a way to reuse attribute content as metadata in the way we reuse element content

Ben: propose to respond that RDF/A can annotation forms elements

Mark: and further note that reuse of attribute content for metadata will add too much complication

RDF Schema for rel attribute

DanBri: it's been some months since I last looked at it
... I recall that it felt done except for re-doing the image
... and my knowledge of the specific link types in HTML was sketchy
... current markup was done by hand yet the text is almost the same as is in the XHTML spec; can we find a way to extract this from the XHTML spec?

Ben: does this seem like something that should be integrated into the XHTML2 spec?

Steven: we've already defined the meaning of the attribute values
... might be no objection to including an RDF schema for the values as an appendix

DanBri: http://www.w3.org/2005/05/hrel/ does have a list of open issues
... in some cases I tweaked the [HTML] text to read better as a property description

ACTION: All in the TF to look at http://www.w3.org/2005/05/hrel/ to decide whether it's ready for WG review [recorded in http://www.w3.org/2006/01/17-swbp-minutes#action03]

Next Meeting

Ben: from next week our telecon time is 1 hour earlier

next meeting: 24 Jan, 1400 UTC

ACTION: Ben draft a response to Gary Ng's comments [recorded in http://www.w3.org/2006/01/17-swbp-minutes#action04]

ACTION: Ben to draft a new example of RDF/A as an XHTML document that is its own RSS feed [recorded in http://www.w3.org/2006/01/17-swbp-minutes#action05]

Ben: should we ask the WG to publish the Primer as a Note or as a WD?

Ralph: I propose WD -- that we set an expectation that we intend to update it

RESOLUTION: we ask the WG to approve publication of the RDF/A primer as a Working Draft

[adjourned]

(scribe note: the following were copied from the previous meeting record. These were not explicitly discussed during this meeting.)

RDF-in-XHTML TF

17 Jan 2006

Attendees