See also: IRC log
ACTION: Ben summarize the question of semantics of the HTML role attribute and solicit feedback from RDF experts [recorded in http://www.w3.org/2005/10/04-swbp-minutes.html#action01] [CONTINUES]
-> Re: The ROLE attribute [Lisa 2005-10-03]
Lisa: WAI PF is using Role to help with inaccessibility to avoid guidelines such as "don't use javascript"; e.g. tell people how to use javascript in an accessible manner. Role is a very basic concept from an RDF perspective; e.g. a Submit button is a kind of a button which is a kind of widget which is a kind of roleso a DIV or any other element, e.g. in SVG, can be given a Role and the accessibility module will know what is expected. We're also working on accessible properties such as State -- knowing what State an object currently has and setting this state. We're soon going to be addressing structural issues. There is some basic structure now; e.g. a spreadsheet is a kind of table which is a kind of grid -- to be able to adapt a page according to the structure of its role. The third use case will define different types of content; e.g. sitemap, to help people with learning disabilities; can adapt the page to a form that works for the person according to its content role
Ben: is role something that helps describe to the browser the purpose of the UI element?
Lisa: yes, purpose but also what is expected; e.g. if role is checkbox then I can expect the states are 'checked' and 'unchecked'
-> 25. XHTML Role Access Module
Ralph: what is WAI PF WG interpretation of the reference to the about attribute in 25.2?
Lisa: not sure; WAI PF has been assuming that we can use Role for our purposes; this element has a role of a certain type; e.g. an element is an instance of a Class. If you have basic roles we hope it will be possible to extend those roles. W. Do people publish new role definitions? Do document authors override these?
Ben: it appears that WAI PF Role is a subclass of dc:Description?
Lisa: no, we're just requiring that Role have at least one dc:Description
Ralph: I wondered at first why Role was an OWL:Class until I saw that you used OWL Restrictions
Mark: I think we've agreed that role does create a triple. It's a separate question whether that triple has rdf:type as its predicate. I don't see that the answer to this second question affects whether WAI PF can get what it wants. Why does PF subclass everything from Role? why inherit everything from this relationship?
Lisa: I first wanted to be sure that people were aware of the WAI PF work but yes, I think you're probably right that the decision [about rdf:type] won't affect us. We have overloaded the word 'role'; think of this as WAI:role -- we're making instances of Classes. Predicates will refer to a separate document of states. We're not thinking of this role as being an attribute but rather a set of values
Mark: but something _plays_ a role whereas [we are thinking that] role is a predicate, so a toolbar plays a role, it is not an instance of a role
Lisa: we're thinking of role in a different way; we don't use it as a predicate
Mark: you could say that a property of a toolbar is that it plays a WAI Role but to say that it _is_ a WAI Role seems wrong. Ehat's the connection between WAI role and XHTML2 role? It doesn't feel correct to me that a tri-state checkbox is a kind of role
Lisa: I think we're in agreement; we're using the word 'role' in 2 ways. We're using 'role' as a base class whereas it's commonly used as a predicate
Mark: given a tweak in the words, then, I think the XHTML2 use of role is exactly consistent with what WAI wants
Jeremy: I think the relationship between xhtml2:role and rdf:type is a red herring. Resources have a role (which might be better pronounced as "hasRole"). We can think of these as classes of resources. this might be rdf:type or not. subject is an xml element, predicate is hasRole, the object feels like a class to me. I'm indifferent to whether hasRole is a subProperty of rdf:type; this decision involves some idea of the information modelling so it's appropriate to listen to Lisa explain WAI PF's modelling but the definition of hasRole is independent of RDF/A. It's a real decision but not one that we need to make for RDF/A
Ralph: I'm with Jeremy on this. It is in our scope to determine the meaning of xhtml2:role. A question for HTML WG and this Task Force is whether the way the 25 August WAI PF document uses xhtml2:role is consistent with our expected semantics
Ben: we can punt the question of whether xhtml2:role is a subProperty of rdf:type but we still need to decide the meaning of xtml2:role
Jeremy: listening to Steven, I hear him proposing that xhtml2:role is subProperty of rdf:type. I'm happy to defer to [HTML WG] on that question. I would only object if xhtml2:role was considered to be an alternate spelling of rdf:type. A semantic relation between the two should be reflected in the schema
Mark: I think WAI PF has only started to realize all the issues. It may be an issue if we infer rdf:type from xhtml2:role as this would cause WAI Roles to get many properties that they did not intend. I worry that we'd be inferring all sorts of wrong things if we say "this DIV _is_ a toolbar" rather than "this DIV plays the role of a toolbar"
Jeremy: regarding the document as an XHTML 2 thing, the id refers to a fragment of the document. In the RDF graph we're thinking in a more abstract way. It seems plausible to me that one could say that section of the document is a toolbar. How much does working with the fragment viewed from an RDF perspective allow us a different view than looked at from an HTML perspective?
Mark: my recent example making a foaf:Document be an instance of a foaf:Person, to which DanBri objected, made me concerned about this modelling of role
Jeremy: these kinds of things are uncovered whenever you drill down into an information model. If we decide that there is no relationship between xtml2:role and rdf:type but let WAI define one if it needs it, there may be harmful interoperability problems if other groups also want to define such a relationship
ACTION: Mark and Ben to check edge cases of inheritance in RDF/A [recorded in http://www.w3.org/2005/07/26-swbp-minutes.html#action06] [CONTINUES]
ACTION: Ben to put together the "ACID" test for XHTML2 RDF/A [recorded in http://www.w3.org/2005/07/26-swbp-minutes.html#action02] [CONTINUES]
Ben: these two actions merge
Ralph Guus would like us to have a document published ahead of time to give the WG. In a perfect world, this document would be a new specification of RDF/A
Jeremy: we want WG input on some of the isues we've decided, e.g. CURI
Ralph: I disagree that we've decided CURI. I have raised the CURI idea to my colleagues within the staff and there's been some significant amount of pushback against it
Jeremy: feedback in HP has not been negative
ACTION: Mark write CURI specification by 10 Oct [recorded in http://www.w3.org/2005/10/04-swbp-minutes.html#action02] [CONTINUES]
ACTION: Ben put notes in the Web from Boston discussion with Mark [recorded in http://www.w3.org/2005/10/11-swbp-minutes.html#action05]
ACTION: Ralph and Ben to augment the issues list [recorded in http://www.w3.org/2005/09/27-swbp-irc#T14-30-04] [CONTINUES]
Ralph continues, with some progress
Ralph: do we expect to have a new document specifying RDF/A before the f2f?
Ben: yes, I think that is important
Jeremy: I have quite a few deadlines at the beginning of November, so I may not be able to have the revised XSLT until after the f2f but I will review a document and give some idea of implementability before then
[adjourned]
Change Log:
$Log: 11-swbp-minutes.html,v $ Revision 1.2 2005/10/14 18:08:56 swick Cleanup for publication