UWA Teleconference
25 Oct 2007

Publishing last weeks minutes

Any objections?

None

Recap of Mobile Ajax Workshop

<Dave> Given that Rhys isn't here I propose that we defer this until the face to face.

<Rotan> Yes, it would be a good thing to cover in the joint UWA/DD discussions on Thursday evening.

<Dave> okay, let's do that.

Reviewing XHTML Role and WAF Widgets

<Dave> We have been asked to review the latest drafts for these two specs. In the absence of volunteers, I will try to draft something for us to look at in email. XHTML2 WG would like to have feedback before the f2f

ACTION: Dave to draft reviews for XHTML Role and WAF Widgets

* trackbot-ng noticed an ACTION. Trying to create it.
RRSAgent records action 1
Created ACTION-62 - Draft reviews for XHTML Role and WAF Widgets [on Dave Raggett - due 2007-11-01].

Layered UI

<Dave>I have been reading a report by Butler group (DataMonitor) on rich web applications. It is clear that there's a gap in the market for standards-based declarative authoring solutions for rich web apps. JustSystems has an example close to this called XFY, and DIAL represents another. We need to be much more proactive on explaining the business drivers for using XML for declarative authoring of rich web apps.

<Dave> recaps his email:

Rationale

Declarative authoring techniques together with the appropriate tools can dramatically reduce costs and improve security, quality, accessibility and ease of delivery to multiple channels. Standards are needed to ensure interoperability across authoring tools and to mitigate customer fears of being locked to particular vendors.

Layered UI Architecture

This is borrowed from academic work on more flexible approaches for declarative models of user interfaces. There are five layers:

1. Application tasks, data and meta-data
2. Dialogue models (and history)
3. Abstract UI (modality and device independent)
4. Concrete UI (modality and device dependent)
5. Physical realisation for specific devices

Requirements

These are some proposals for key requirements that the architecture must satisfy:

• XML-based UI markup with lots of existing work to leverage, e.g. XUL, XAML, MXML, Morfeo, OpenLazlo, UIML, UIX, xfy ...
• Use of XML for constraint-based layout and UI attributes
• Must support UI skins that change appearance and behaviour, e.g. as for XMMS and WinAMP
• Declarative event handlers that can raise further events, e.g. to invoke local and remote services (integration with SOA), and to generate higher level (intent-based) events from lower ones.
• Rich dialogue models e.g. event driven state machines (SCXML)
• Must enable high level authoring tools for non-programmers
• Must be extensible using templates and associated resources provided by programmers
• Must support distributed model where different layers may reside on different machines with event flows across NAT and Firewalls, etc.
• Declarative mapping between layers (e.g. XSLT)

The physical realisation can be messy as it has to deal with the wide variations in device capabilities and associated interoperability problems with HTML browsers, Flash, Silverlight, JavaME, .NET, etc. Mapping to physical layer is guided by declarative policies provided by authors, but is essentially a compilation process.

Links to further details

• XUL
• XAML.net
• Morpheo
• Sliverlight
• MXML
• Open Lazlo
• XML UI Workshop
• UIX
• UIML
• XAML

<Dave> invites Jose to give his presentation.

Jose speeds through the first slides.

<Rotan> I am holding onto a question for Jose regarding XHTML Role, which could be relevant to his presentation.

<Rotan> And also another question re slide 21: the use of simple layouts like grid/horiz/vert.

<Rotan> Think we should look at the GridBag layout. It's a very useful general layout, and I have an XSLT to map from hierarchical layouts to the equivalent <table> markup.

<Rotan> I'd be happy to donate it :)

Slide 12: left side is abstract user interface. Going lower gets more concrete.

Dave: Questions during the talk, or at the end? Jose: Better to wait until end.

Slide 12: Left-hand side, different languages or technologies that might be used in each layer. Will need technologies for mapping between the layers and for defining the policies.

<Jose> Questions on general problem?

<Dave> In the email had an additional layer above the top. The idea there is to have an ontology/description of kinds of data (metadata), eg: fields required, contextual dependencies, naming of tasks, etc. The dialog level would be below that, and might sometimes have need for history. Lower level example: a double-click requires the need to compute time from previous click. Or, where you need to combine multiple events like mouse gesture and spoken input .

Also, eventing. The role of events at different levels in the architecture.

Concrete events get mapped into higher level events at the abstract ui layer.

<Jose> Could there be work to redraw the picture in slide #12 to reflect his?

<Dave> Its okay for today, as this is an evolving story.

<Jose> Will present more complete picture at the face to face.

* Presentation continues at slide 18

Slide 26: link to wiki page to an idea for separating DISelect

* Presentation complete

<Rotan> Typing some ideas into IRC. Absence of a mechanism for abstracting... XHTML Role attribute might be a means of capturing missing metadata. Combined without a layout mechansim (GridBag). Mapping roles to GridBag might provide a simple mechansim.

<Jose> Have not thought about that. Don't know if role attribute is the best way to deal with layout.

<Rotan> Would have a mapping.

<Jose> Using CSS to assign a layout to an element you only use the role attribute.

<Rotan> CSS could reference the role definitions and apply a layout based on the roles.

<Jose> yes, that could be a mechanism for mapping between elements and layout. But, the layout would be ... ?

<Rotan> Could be using CSS to place the contents or could use a mail-merge approach where you have a template document and use roles to place appropriately to provide adapted content.

<Dave> I think that XSLT and XSL-FO represent interesting technologies and that it will be worth holding some joint discussions on requirements and opportunities with the XSL WG

<Rotan> CSS is a moving target. Prefer the second option of defining a layout mechanism indep. of CSS.

<Dave> I am planning on a meeting with XSL WG team contacts (Michael and Liam) at the TP.

<Rotan> Thinking... by combining XInclude, an XHTML template, DISelect for contextual selection and a source document using XHTML Role, you could build a simple process to apply contextually sensitive layouts to content.

<Dave> Indeed.

The XSL WG is currently considering adding constraint-based layout to XSL-FO, and in principle that could be compiled into HTML+CSS for the physical realization.

<Dave>Thanks Jose for his presentation and Keith for minuting. We will talk about layered UI further at the F2F, including possible changes to the charter to ensure we can proceed to specification work.

<Dave> We won't meet next week, which simplifies things due to Europe switching clocks one week before North America.

<Rotan> See you in Boston. :)