• RDFinXHTML Taskforce Issues

RDF in XHTML Taskforce Issues

This version
http://www.w3.org/2006/07/SWD/wiki/RDFinXHTML_Taskforce_Issues

Previous version
http://www.w3.org/2001/sw/BestPractices/HTML/2005-current-issues

---

Index

---

This page summarizes open and resolved (or closed/deferred) issues of the RDF-in-XHTML taskforce w.r.t. RDFa.

1. Open Issues

This sections contains issues that are pending and need to be addressed by the TF.

1.1. Reification (was: No. 2.)

The concept of reification has been confusing and awkward to use (in RDF/XML at least).

1.2. Custom Attributes for RDF shorthand (was: No. 3.)

1.3. Role Attribute for RDF type declarations (was: No. 4.)

Initial Motivation: to reduce the use of rdf:type in some examples.

Discussions:

• July 2006: http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2006Jun/0009

• August 2005: http://lists.w3.org/Archives/Public/public-rdf-in-xhtml-tf/2005Aug/0003

• July 2005: http://www.w3.org/2005/07/26-swbp-minutes.html#item02

Points of Debate about role:

• should we use rolein some RDF way that is equivalent to rdf:type at all?

• if we do, does that mean role is equivalent to rdf:type, or is a subproperty of rdf:type?

• is role more of a declaration of "plays the role of" for purposes of client user interface?

• maybe we need a new syntactic sugar for rdf:type? But that's possibly a slippery slope.

Points of Debate about class:

• should class be rdf:type? Another predicate?

• see Mark's email.

1.4. (Local) blank node identifiers (was: No. 5.)

How to refer to blank nodes (nodes that do not have an externally referencable identifer) from within an XHTML document. See how do we get anonymous nodes?.

TENTATIVELY RESOLVED: blank nodes are used in striping mode, where they can be the object of a single statement, and the subject of a number of statements.

1.5. CURIEs in Predicate Attributes (was: No. 12.)

Currently, about and href attributes are of type CURIE/URI, which means that the value of the attribute can be either a CURIE or a URI. When it is a CURIE, it needs to be wrapped in square brackets as follows: [cc:license], in order to differentiate CURIEs from URIs.

Should rel, rev, property have the same CURIE/URI type? Or should they be CURIE only, since there is no precedent for them being URIs?

The advantage of CURIE only is that we get automatic backwards compatibility for things like rel="next", which would be interpreted as CURIE in the default XML namespace of xhtml2.

If we go with CURIE/URI, then we can always declare parser-level legacy conditions for things like rel="next", which would immediately be transformed into rel="[next]" or maybe even more directly to rel="[xhtml2:next]".

1.6. xml:lang (was: No. 14.)

In the RDFa Syntax document (as of 2006-12-19), xml:lang is discarded for XML literals. While this may have been okay for RDF/XML, it's much worse for XHTML, which should be treated with "more respect." The fix for this may involve adding an extra SPAN or DIV as appropriate to XML literals.

1.7. Comments from Christian Hoertnagl (was: No. 15.)

We need to address all comments from Christian Hoertnagl.

1.8. RDFa Containers: Addressing DL (was: No. 16.)

The RDFa containers proposal does not address DL. We should eventually address it.

1.9. Nested meta's and link's (was: No. 18.)

What happens if we nest meta's and link's ?

Proposed Resolution: the subject of the inners are the same as the subject of the outer. In particular, if the outer has a about, then that's the subject for the inners. If the outer has no about, then the subject is the outer's immediate parent. A good example is found here.

1.10. XHTML 1.1 Namespace doesn't end in / or # (was: No. 19.)

How can we resolve CURIEs?

Note (2006-05-13): As the TF is considering grounding CURIEs in the current URI (possibly followed by "#"), this may no longer be an issue.

ToDo (Ben, please clearify): should we close this one?

1.11. The RDFa syntax document (was: No. 23.)

The RDFa Syntax document has some bugs and is out of date. This is a non-inclusive list:

• 6.2 - the geo:lat should be an XMLliteral (EliasT)

• TBC

1.12. EliasT and LeeF comments (was: No. 24.)

• EliasT sent comments on June 3rd, 2006 regarding the RDFa Syntax document.

• LeeF sent comments on May 31st, 2006 regarding the RDFa Primer document.

These comment should be addressed ASAP

ToDo (Ben, please clearify): should we merge this one and the previous one (was: 23)?

2. Resolved, Closed, or Deferred Issues

This sections contains issues that have been either resolved, closed or deferred due to mail discussions or telecons.

2.1. QNames in href and about (was: No. 1.)

There is strong desire (or requirement) for using QNames in place of URIs in order to achieve compactness of an HTML document instance. One use case expects to have substantial numbers of such URI references within a single document and wants to save storage space and transmission time.

• Option A: qhref and qabout parallel href and about with datatype QName. See 2005-07-12 meeting record

• Option B: a new syntax for distinguishing a QName from a URI; compact URI. See Mark Birbeck message of 2005-07-19 and Jeremy Carroll's summary of qname/bnode proposal.

2.2. Applying Metadata to the src URI (was: No. 6.)

Images and Objects have a src attribute. There is currently no way to attach metadata to the URI value of that attribute without repeating the URI in an about attribute. Should we find a way to change this such that the image or object URI is not repeated?

A few options:

• do nothing, you must repeat the URI

• add some link and meta statements within the img tag (which is no longer empty in XHTML2), along with some conditions that state that the src attribute functions like an about attribute for all children RDFa statements.

• something else?

ToDo (Ben, please clearify): TENTATIVELY CHANGED to NO: src should behave like href for RDFa, it's an object, not a subject. See telecon notes.

2.3. Syntactic Sugar for class attribute (was: No. 7.)

Should the class attribute be syntactic sugar for rdf:type? Consider the following XHTML:

<div class="biblio:Publication">
   <meta property="dc:title">A Standards-Based Virtual Machine</meta>
</div>

which would, under this new syntactic sugar, provide the following triples:

_:div0 rdf:type biblio:Publication .
_:div0 dc:title "A Standards-Based Virtual Machine" . 

The class attribute would allow for multiple values, just like the current definition. Each value would be the object of a new triple.

2.4. Reifying Any Statement, not just link or meta (was: No. 8.)

We may want to reify any existing RDFa statement, not just those expressed with link or meta. In fact, we can do this without complicating the processor at all. Simply, when trying to determine the subject or object of an RDFa element by traveling up the DOM tree (or checking the immediate parent in the case of a link or meta), if an element with a rev,rel, or property attribute is encountered first, then the child RDFa statement has, as subject or object, that reified parent statement.

For example:

This document is licensed under a 
<a about="" rel="cc:license" href="http://creativecommons.org/licenses/by-nc-nd/2.5/">
  Creative Commons License
  <meta property="dc:creator" content="Ben Adida" />
</a> 

would yield the following:

[ <> cc:license <http://creativecommons.org/licenses/by-nc-nd/2.5/> ]
  dc:creator "Ben Adida" . 

2.5. Making link content clickable (was: No. 9.)

Should a link element be allowed to have content which would then be clickable? The reason for this is that link plays a special role, e.g. in reification and blank node annotation, and it doesn't seem like a good idea to exclude clickable links from those special roles.

For example, assuming one wants to say that "Fyodor Dostoyevsky" is the author of a quotation, one could write the following statement:

<blockquote>
       <p>
            Rodion Romanovitch! My dear friend! If you go on in this way
            you will go mad, I am positive! Drink, pray, if only a few drops!
       </p>
       -- <link rel="dc:creator" href="http://dostoyevsky.com">Fyodor Dostoyevsky</link>
</blockquote> 

2.6. Predicate Inheritance (was: No. 10.)

It may be possible to reintroduce predicate inheritance without exploding the complexity of the processor. The way to limit this feature while maintaining its advantage is to mandate that a predicate is always inherited along with its subject, and that children elements cannot override the subject and expect to inherit the predicate.

Here's how it's useful:

<div about="#publication_2" id="publication_2" class="biblio:Publication">
  <link rel="dc:creator" href="http://www.blogger.com/profile/1109404" />
  <meta property="dc:title">XForms and Internet Applications</meta>
  Tags:
  <ul rel="taxo:topics">
    <li href="#tag_standards">Standards</li>
    <li href="#tag_xforms">XForms</li>
  </ul>
</div>

Note how the ul element clearly inherits the about from the div, which is then the subject of all triples with predicate taxo:topics. The result is:

<#publication_2> rdf:type biblio:Publication .
<#publication_2> dc:creator <http://www.blogger.com/profile/1109404>
<#publication_2> dc:title "XForms and Internet Applications" .
<#publication_2> taxo:topics <#tag_standards> .
<#publication_2> taxo:topics <#tag_xforms> . 

The general rule is that an RDFa element without a predicate will search up the DOM tree for the closest predicate-containing element. That predicate-containing element cannot have its own object within the same element. Subject resolution then begins on that predicate-containing element.

2.7. CURIE Base (was: No. 11.)

If a CURIE is unqualified, what is its base? It could be either the current URI base, or the default namespace.

2.8. Plain Literals (was: No. 13.)

Given the RDFa statement:

<meta about="" property="dc:creator"><b>B</b>en <b>A</b>adida</meta> 

How might one designate that the value of the object should be the plain literal "Ben Adida"? Should there be some means of instructing the parser to concatenate all the text elements of children elements?

2.9. head edge case: default about (was: No. 17.)

What happens if the head of an HTML document has an explicit about attribute?

Proposed Resolution: the default subject for the whole document is then the value of that about. This is particularly useful for those web pages that describe a non-information resource.

2.10. rel,rev,property is CURIE, or CURIEs (was: No. 20.)

2.11. hGRDDL as a general way to transform HTML into HTML with RDFa (was: No. 21.)

The TF needs to cooperate with the GRDDL folks to think about the hGRDDL proposal.

In addition to this proposal, the TF is considering default hGRDDL profiles for XHTML1 and XHTML2, which would resolve issues such as rel="next". This may require GRDDL to think about pipelining for multiple transformations.

ToDo (Ben, please clearify): Should this one be an OPEN issues, rather?

2.12. Section 3 of the Primer should be better motivated (was: No. 22.)

Gary Ng says that Section 3 of the Primer should be better motivated, a bit more like section 2.

ToDo (Ben, please clearify): Unclear about status.

---

CategoryRDFa