W3C Workshop on Declarative Models of Distributed Web Applications

5 Jun 2007


See also: IRC log


Chaals, Gorm, Marie-Claire, Dave, Steven, Kevin, Rotan, Rhys, Hideki, JoseC, BruceL, CharlieW, MikkoH, Lasse, KoitiHasida, StephaneBoyera, SaileshSathish, NachoMarin, PetriVuorimaa, CorrieTimmons?, FabioPaternò, KangchanLee, JoshueO'Connor, JariKleimola, Joerg_Heuer, GormEriksen, Marie-ClaireForgue
Dave Raggett, Kevin Smith
Steven, steph, Krcsmith, Kevin, Chaals




<stef_dub> draft program: http://www.w3.org/2007/02/dmdwa-ws/program.html

<Steven> (22 present in room)

<Steven> Rotan: Welcome

<Steven> ???, Irish State Development: Welcome

<Steven> Scribe: Steven

Dave: Capturing intentions as a means to reduce costs and improve quality
... Talks are in order to spark discussion
... presentations will be made public after the talk
... please give them to me on this memory stick
... I will publish

(shows agenda)

<rigo> Steven, please scribe here as I will look at it

<rigo> :) thanks

Dave: Longer discussion tomorrow, trying to discover where W3C should be working on in this area

Kevin: Welcome
... I want to set the scene
... workshop title is a bit technical
... I work for Vodafone R&D
... We have an access revolution
... web apps are very successful
... work corss-platform
... can add functionality from the server side
... How should we evolve?

<scribe> ... New considerations include new network possibilities like Wifi, Digital broadcast etc.

UNKNOWN_SPEAKER: and lots of new devices
... in new delivery contexts
... (which captures variables such as location, time, connection rate and so on)
... This leads to new challenges
... like user experience (try looking at a current web page on a mobile phone)
... we need a functional user experience (even if we can't always get an optimum UE)
... There are issues of safety and privacy
... issues of state, like restarting an app from where you left off
... Development for multiple platforms
... and how to overcome inertia
... Standards initiatives
... include device independent authoring
... security
... management of state
... capturing intent
... which is exactly what declarative approaches do: captures intent
... Standards toolkit inlcudes:
... xmlhttprequest
... rex
... dial+xforms
... XBL
... privacy and security stuff
... Use cases:
... watch tv at home and restart watching from your mobile
... share web content with friends
... [scribe missed last point]
... A declarative approach offers a foundation for network-network comms
... eg wifi and rfid for tracking hospital patients
... agents monitoring nutrients in soil and ordering fertiliser
... tracking cars via GPS, and reporting errors
... (and more this workshop)

Dave: Introductions

Rotan Hanrahan, Mobile Aware

Rotan: chair of device independence WG at W3C
... our company delivers content to multiple devices

Hideki Hira, Just Systems

Steven Pemberton, W3C and CWI

Rhys Lewis (Volantis Systems)

Rhys: Chairs and has chaired various groups
... also on the TAG
... we belive in write once and deliver anywhere

Jose Cantera, Telfonica

Jose: I work in open source implementations, interested in declarative approaches for user interfaces

Bruce Lucas, IBM

Charlie Wiecha, IBM Research

Charlie: For me the top order bit is thinking of an app as an end-to-end solution

Mikko Honkala, Nokia Research

Mikko: Worked in XForms in the past
... I want a complete end-to-end model, so that you don't need many paradigms, many languages

Lassa Vaienen (?): Nokia

<Sailesh> Lasse Pajunen, Nokia Research Center

Hasida Kiosha (?)

[scribe misses company]

Hasida: I will discuss constraint systerms

Stephane Boera, W3C

Sailesh Sathish, Nokia Research

<stef_dub> s/boera/Boyera/

Sailesh Sathish: I will be presenting today, I want to hear from you what is feasible

[Nacho Marin] CTIC Spain

<Sailesh> Sailesh Sathish, Nokia Research

Petri Vuorimaa (Helsinki University of Technology)

Corrie Timmons (?)

Josh O'Connor, National Council for the Blind

Josh: Interest in accessible interfaces

Fabio Paternò (ISTI - CNR)

Fabio: Interest in model-based design

Kangchan Lee,

W3C Korea Office

Jari Kleimola HUT

<mikko_honkala> Jari Kleimola, HUT

Joerg Heuer, Siemens

[name4], Siemens

[name4]: Work with 'strange' devices, challenge of integrating UIs on all those devices, without writing a new UI for each one

Charles McCathieNevile, Opera

Chaals: Chair W3C Web API WG
... interested in accessibility
... Opera makes browsers for all sorts of devices
... wedon't want to throw away what people already know
... vast investment in current skillset
... don't throw that away
... it's not that we don't want to take on new technology
... but
... we want to increase compatibility for a migration path

[Gorm Eriksen], [???]

[Gorm Eriksen]: I am also in Web API WG

scribe: I believe in using the technology we already have

Marie-Claire Forgue, W3C

MCF: In charge of mobiweb communications

<stef_dub> Scribe: steph

<stef_dub> Scribenick: steph_dub

Steven pemberton - Loading the silver bullet

<Rotan> Interesting point about spreadsheets being declarative, and obviously easy to use. Must remember that one.

<Rotan> And that's 200k of *compressed* JavaScript in GMaps!

<stef_dub> back

Steven Pemberton - Why declarative approaches matter

<stef_dub> Rotan: google map example ? 200k compressed ?

<stef_dub> Steven: yes compressed

Fabio Paterno - Migratory interfaces

<stef_dub> Rotan: structure of the original apps: do you think it is better to have this structure explicit by apps programmers or the reverse engineer is beeter ?

<stef_dub> Fabio: here we did reverse engineer, to capture the original developer goals

<stef_dub> Rotan: so structural guess work right ?

<stef_dub> Fabio: yes

<stef_dub> Rotan: can the user give input for the guess ? could it help ?

<stef_dub> Fabio: interesting idea, but not easy to support. End-user may not have the right background to give input, but interesting

<stef_dub> Steven: very good point for the declarative approach: if the guess disappear it would help and make better result

<stef_dub> ???: suggestion that the user could give input: it could be wrong also, and give bad feedback. The constraint have to be correct

<stef_dub> Dave: any idea about a good authoring environment that could help capturing hte intent of the apps developper ?

<stef_dub> Fabio: we have an authoring environment where people can enter the logical description

<stef_dub> Dave: missing good authoring tools, so we should define what such a good tool should represent

<Steven> For the (IRC) record - we had a network failure, which has now been resolved

<Steven> Steven's slides

<chaals> Bloody SAS can't get a flight time right :(

<Dave> the remote control of microphone level is a fine example of a ubiweb capability

<Dave> also the ability to set different speaker volume levels for front and back of the room

<jo-siemens> In the back of the room it's still a bit too loud and some feedback reduces audibility

<Dave> blackboards would be useful for events, e.g. in EMMA that represent interpretations of user input

<jo-siemens> Volume is better now, thanks!

<marie> yes, thanks!

<Steven> Mikko Honkala, Nokia Research, Connecting XForms to Databases

<Dave> I coded a spreadsheet in prolog including topological sort of dependencies back in mid eighties on a unix portable that looked like a sewing machine.

<Dave> there is a bar/restaurant across the street, but it might take quite a while

<Rotan> Not sure if the Natural History Museum (5 mins up the road) is open. Free entry. There's a piece of moon rock on the third floor. And a dodo skeleton.

<Rotan> The XFormsDB markup looks complex. Wonder if this would be generated by an authoring/app.dev. tool...

<marie> wow! to the dodo skeleton :)

<Krcsmith> Scribe: Krcsmith

Distributed Media Center - Jari Kleimola and Petri Vuorimaa

Jari: Home media and declarative techniques
... MVC approach
...Problem: multiple remote devices, how to synch and control them
... a remote hub would help
... For multiple homes (e.g. summer cottage, or friends) how to access and share
... (shows DMC architecture diagram, ref: slides)
... (XML over HTTP, distrubuted DOM model)
... Abstract UI transformed into Concrete UI by XSLT
... Markup for various clients (XHTML, MIDPML etc.)
... Can further distribute the UI (e.g. send from mobile to another client)
... Dynamic infrastructure DOM. When a new device indicates presence in network it is added to InfraDOM by REST or REX
... State variables are captured
... UI fragments can be updated without loading whole page (XMLHTTPRequest xhr binding)
... uses a micromodel binding with XPath
... (shows interaction loop diagram)
... User can submit a target (e.g. to another room)
... The 'room' is registered as a listener and is updated accordingly.
... Only 58 lines of markup (ref: XPROC slide)
... XProc is a W3C WD, pipelines of compound steps, micro-operations
... (shows Techniques slide)

Questions: -

Steven: what is it that makes declarative easy to debug

Dave: plugins to CSS
... XBL? Binding to components according Delivery Context

Jari: Not had time yet :)

Rhys: Why is XPROC declarative?

Jari: XPROC handles XML fragments which are declarative. A subset of Xproc was used (although there is no implementation yet)

Dave: Want to be able to reduce power (least power)

Steven: Declarative and Turing complete are orthogonal issues

Sailesh Satish- Declarative Models in multi service smart space environments

Sailesh: We have proprietary implementations in this space but want to bring a more Web-centric view.
... Smart space, "a smart/intelligent environment with which users can interact"
... [serious feedback from mic!]
... Can be logical as well as physical
... (talk through differences to distributed systems ref: Introduction slide)
... support heterogenity.
... Use case: I come to a smartspace with a mobile device. I access a music download environment. I should be able to use the dial and

buttons on my home stereo to navigate through the mobile-browser displayed song list (i.e., use the physical controls from another

hardware unit to interface with the web application)
...infrastructure: blackboard design pattern. Blackboard can be accessed by devices. Sensors and devices can put info there, listeners

can access this data

scribe: How can we do this with a Web-centric approach?
... Delivery Contect Client Interface provides DOM based API for static/dynamic context data access and provision
... Composite DCCI: individual delivery context client API across multiple clients (analog to a super-blackboard)
... REX is a candidate for a synchronisation protocol.
... DCCI to perform client side, runtime adaptation
... Need a process for a standardized provider interface
... Service invocations through DCCI is a missing piece
... Need careful management of metadata when synchronsing nodes
... Need an access point to identify that you are entering a composite model (the smart space blackboard)
... DCCI is object centric, blackboard is data centric. So no interfaces to an exact data structure in DCCI yet. e.g.Type information in


scribe: State machine approach. Smart Space applications can be seen as state machines (adaptation based on environmental conditions).
... How to expose device-specific code to applications? e.g. through ontologies/vocabularies. Invoked through conversational approach.
... SCXML is one such candidate
... Security and privacy are issues!


Rotan: Fascinating as you are part of an environment not a machine. What happens if you are in more than one smart space

Sailesh: This has been a problem
... when there are overlaps we try to address the models as separate.
... So the data is distinct. Object-centric, can have metadata embedded which indicates which smartspace(s) it belongs to.

Rotan: e.g. ambient audio vs direct audio

Sailesh: Every environment has one smartspace

Rotan: But logical smartspaces are not physical?

Sailesh: Not necessarily, it is a unified space

Rotan: A portable smartspace (e.g. my Personal smartspace)

Sailesh: Smartspace on a device is always yours to access, and can come into contact with others.
... you can disambiguate

Steven: Problems with blackboards?

Sailesh: We take object (not data based) approach helps scalability problems

Dave: Multi-modal work (EMMA) makes use of blackboard approach (record of what was said)

Rhys: Blackboard as semantic event listener (sharing mechanism).
... What do you see as missing?

Sailesh: RE has to be modifed by itself. How to communicate reduncant node information etc needs to be addressed.

Connecting XFORMS to databases, Mikko Honkola

Mikko: (presents slideset)
... What do we mean by declarative? is a good question
... Possible to separate out parts of a language which are not Turing complete
... Even simple web authoring is difficult
... Spreadsheets allow declarative techniques for non-programmers
... (ref: problem statement slide)
... Business logic is very divided
... Binding the form to a database allowing multi-user concurrent access to the data (NB form in an AJAX, frequently updated submission)
... XFormsDB as a superset of XFOMRS 1.1 + CSS 2.1
... XForms client is not required (fallback to current browser markup)
... Current syntax is proof of concept
... (ref: example update slide)
... Pointing to subtree in db, generating fragment transferred to the client.
... Synchronisation to take care of multiple concurrent accesses
... (ref: synchronisation slide)
... using well-known 3-way synchronisation algorithm.
... Full transaction support needed for future (e.g. moving between different databases)
... (ref: Architecture)
... AUthoring tool not implemented yet.
... Query definitions are stripped leaving XForms 1.1
... When requested XForms is translated into AJAX
... Relational database can be integrated with XML mapping
... XFormsDB blog as a prototype
... Can use SQL datasource as well


Steven: Advantages over encoding into the URL?

Mikko: Differences are just in the syntax

Steven: Malicious clients?

Mikko: Possibly more security with our approach.

Rotan: Markup is complex, and tools support is important. Any prototypes, issues?

Mikko: Out of scope of this project. Extensions are quite small. We could reuse an open source XFORMS editor.

Rotan: Storing app data as XML data. Can you store Efficient XML?

Mikko: You could insert this step.

Rotan: Recommended to build on existing work to reduce overhead.

?: Security

Mikko: Can be added and scaled

Dave: Declarative approaches should support this work

Mikko: ... Will be interested to see the options.

(Break for lunch)

<Rotan> (Lunch is ending)

<Rotan> "Declarative Models of Distributed Web Applications"

<Rotan> Live demo - http://www.xfy.com/manual/dev/developer/1.4/howto/xvcd_tutorial.html

<Rotan> Also see www.xfytec.com

<Rotan> Note that again there is a need for tooling. WYSIWYG GUI tools needed to support these technologies.

<Rotan> Firefox, eclipse and other tool frameworks help to make such higher level tools easier to build.

<Rotan> WYSIWYG tool hiding underlying detail enables greater use.

<Rotan> Dave: but WYSIWYG can hide richness of the underlying technology.

<Rotan> This technology was originally developed as an office suite.

<Rotan> Wanted pure XML solution. Evolved into the current system.

<Rotan> Technology also handles the backend.

<Rotan> xfy also applies validation all the time.

<Rotan> In fact, it prevents author from creating invalid material.

<Rotan> Speaker - Koiti Hasida

<Rotan> "Compositional Business-Task Organization"

<Rotan> I remember many years ago being taught of the 13 forms of temporal overlap. This could be used in this constraint-based system if timing is part of the reasoning.

<Rotan> This constraint chain reminds me of cell dependencies in spreadsheets.

<Rotan> Sounds like an enhancement to transaction models.

<Rotan> RH: Also has aspects of theorem proving.

<Rotan> KH: Yes. And these CBT systems are easy to automate.

<scribe> Scribe:Kevin

Position Paper Lasse Pajunen (Nokia)

(presents slideset)

Lasse: Challenges in designing web apps
... from client side, how to make friendly UE across multiple devices
... with different capabilities

<mikko_honkala> capabilities are dynamic ... from GPRS to WLAN

Lasse: For dynamic capabilities, the client knows what the value is (bandwidth etc.)
... To face these challenges, we can use CSS (fonts, layouts etc.) and SVG
... Implicit features (speed of page rendering)
... If pages are designed properly implementation can take advantage
... Simplicity very important
... Most web pages authored by GUI designers, not programmers
... So need ways for these authors to do their job properly.
... Classification for UI generation
... Mobile/legacy devices that don't support full HTML, CSS.
... Low capabilities for these devices, how to provide good service?
... Fortunately this category is shrinking as XHTML and CSS support grows
... Normal clients, how to ensure optimal use of bandwidth/memory
... Provide benefits to the powerful clients.
... Modeling interaction for Web applications
... In Web sites you do things in a certain order
... e.g. in a blog, need to discover, register, comment on blog, logout
... All of these actions are linked by a UI process, cannot be done separately
... Need to model the computational logic as well
... Need to bind the UI to the dynamic data in the db, other services
... Must model the whole package including interaction and related data.
... One way is to use Web services and service composition (WS-BPEL and WSDL)
... to describe all the data, computation, presentation services.
... Maybe by treating Client browser as a service it will simplify matters.
... Thus describe whole workflow package in one uniform matter.
... Benefits to service authors
... This can enable easy updates to the workflow if one component changes.
... Allows policies to be implemented.
... (ref: Future work in W3C)
... HTML, CSS support will be available in most devices soon.
... Hands free, usage by the blind are examples of delivery contexts that require adaptation.
... AJAX mechanism can be used for client to poll a proxy for updates
... The problems can be solved with the current solutions.
... Should we change underlying technolgies, or describe best practice on how to use existing technologies?
... My opinion is the latter.


Charlie: do you have to push to the browser? Or can an interaction force that?

Lasse: e.g. Web based Email. First, open app, and if someone has sent a mail you want it to appear.

Charlie: But that is separate to whether it is push or pull.

Lasse: Depends on how tightly we want to integrate the BPEL service.
... What I'm proposing is that we skip UI component and that the BPEL process controls all the components (sends HTML to the browser).

Rhys: Web services important, maybe not integrate fromt-end too tightly.
... Legacy devices only disappearing in Europe/North America.
... So not full support yet for XHTML/CSS.
... Server side adaptation still important.

Lasse: Based on studies around mobile web usage, only people with good browsers are using services.
... So poor/limited browser support does not drive usage.
... More crucial to deploy full web browsers.

Rhys: For Web surfing, I agree, but web application usage (buying ringtones) is a major part and can be supported on legacy devices.
... Agree that it is important to improve XHTML/CSS support.

Dave: SMS support important (Stephane's work for emerging countries)

<Rotan> Developing countries have many needy users who must suffer the old devices. They too deserve Web access.

Collage (Bruce Lucas, IBM)

Bruce: Declarative programming model, Collage.
... There is a mismatch between loosely coupled apps and monolitic programming models.
... (ref: Collage motivation and goals slide)
... Programs built as composition of web components.
... To us, declarative means 'what' not 'how', only declaring what you need, more composition.
... (ref: Outline slide)
... How to use RDF for everything!
... (ref: Data model section)
... (intro to RDF)
... Mutable entities in Collage allowed via composite values
... e.g. a Name and a birthdate of a person
... can be retrived/updated via HTTP GET/POST to URL
... (ref: Execution model)
... (Bruce reads through Execution Model Concepts)
... Cascades are modeled with Bind constructs
... (Bruce goes through example)

<marie> yep, just issued at the top of the hour

Bruce: Flexible decomposition more flexible than subclassing as no knowledge needed of DATE3 class (ref: Flexible decomposition slide)

<Steven> News item

<Steven> Press release

Bruce: (shows demo, including the ability to declare a workflow step that has been inserted into the original bookstore illustration)
... (ref: Device Adaptation slide)
... Shares technology with other layers


Dave: what determines ordering of UI?

Bruce: Don't know yet, maybe constraint based.

Jose: What are main advantages against classical approaches
... what are goals, results etc.

Bruce: End to end uniform model, overcome silos of UI and data model

Charles: Similair to ARIA, have you looked at this?

Charlie: Should look back at this, Collage should be able to do this.

<Rotan> Information on tonight's meal:

<Rotan> This is the driving route to "Break For the Border". Obviously a walk would avoid the one-way traffic system. It's a nice day, so walking should be pleasant.

<Rotan> http://maps.yahoo.com/broadband#mvt=m&q2=53.341098%2C-6.264375&q1=53.341777%2C-6.249097&trf=0&lon=-6.254933&lat=53.341226&mag=3

<Rotan> We have a reservation for W3C at 7:30pm. This is a link to info about the BftB venue.

<Rotan> http://www.dublinpubscene.com/thepubs/breakfortheborder.html

<Rotan> The meal includes starter and main course, and a glass of house wine (or soft drink). Dessert, if you insist :)

<Rotan> For those who would like to sample additional drinks, please note that BftB is also a *pub* (however, we ask that you pay for extra drink yourself).

<Rotan> If you arrive early, the table is booked in my name (Rotan Hanrahan).

<chaals> scribe: Chaals

<scribe> ScribeNick: chaals

Conversation hour

What do we mean by declarative

DSR: Collecting what needs to be done, not how

RH: Would say it is what is to be achieved... e.g. if it is transfer of info from user to system, there is a description of the state you want to reach.
... rather than the mechanism by which that should be achieved

CW: Most task definitions are recursive. So you can decompose it all the way down...

LV: a bad example is "language that has only one word, and the interpreter does what they want". It is about things that the system offers, as opposed to giving commands. We can say CSS is very declarative, but if you have to do explicit commands it is more like a procedure - so declarative is a relative concept, and the higher we go the more declarative we get. E.g. C is more declarative than lower languages, ...

CW: (same speaker) it is like th relaxing of the explicit programming

MH: I don't think declarativeness is a goal in itself. If there is something we cannot do in the current level of abstraction and we can make a higher level of abstraction we should.

<Charlie> * "CW" above is Charlie

DSR: Would like to expand the number of people who can build applications, reduce costs, ...

RL: Pushing up the level of abstraction is important.
... specifying things on the screen is pretty good, but now we want to get some of that for application development and get it generated automatically

LV: If the goal is to make it simpler, you can change the language or provide a library of commonly used functions....
... so in which domains do we need to change the language?

JH: This is about redcing the variability of what it does, so that you can have broad interoperability easily.

RL: The point about reducing power is good. Authors need to stop writing HTML and write XHTML 2 and Xforms which are nice enough for us to change into things that are appropriate for any kind of browser. Decoupling what the author writes from what you send is good for doing that.
... more important is providing pre-canned pieces. You declare the behaviour you want. It constrains what people want to do, but makes the easy things simple (and script allows the complex things to still be possible

CW: Want to hold on to the point about composability.
... having the least powerful stuff, so you have composability, is important.
... it is not *just* about the simplicity of authoring

SP: A magic word is tractability. If there is API support you are half-way there. Having the semantics in the language allows you to deduce what is there more effectively.

LV: Should people use a text editor to create content? If you agree that people should us WYSIWYG editors, what is the difference between HTML and XHTML 2 so we could keep the environment and get authoring tools to move higher up in abstraction. Then XHTML 2 is not abstract enough, but what we should be discussing is what the layer above is, and how to transform it down to HTML...

RL: Problem is HTML isn't sufficient for describing what you are trying to do.

<Rotan> Agree. You at least need XHTML for authoring. (Not to be confused with what you might eventually deliver.)

RL: this is why we ended up going to XHTML 2. There will be people writing with emacs, so we need something that people *can* write. But my desire is to be able t build things based on drawing UML diagrams and create real applications from them
... Everyone says they haven't got to UI yet, so we need a markup language

JH: We have significant models like SVG,HTML, and declarative is sort of model-based. In other areas, like how to navigate a car-based system, it is not so clear how to do it so people use scripting...

<jo-siemens> JH is Joerg Heuer

Petri: Needs to be interoperability beneath the tools that I really use. XHTML is below them - I am looking for something above them...

FP: XHTML is pretty low level, and we need some higher abstractions....
... if we want end users to modify their applications without being programming experts, we need some clearer explanations of the key points...

RL: There has been work in accessibility about the notion of role-based events - describe the intent of the event, rather than the particular mapping to a given device. This used to be called the accesskey problem. You want to bind particular UI interactions to some higher levels than particular input devices, etc...

[this is a problem that is known as being a decade old...]

CW: Focusing on abstraction doesn't stop us having to define markup language. Programs have to analyse and work with the stuff as well.
... we were asked if UML langauge could handle this, and it couldn't, because it didn't have enough in it to do the job...

JOC: Being able to define what to do in a machine-readable way could be useful. Where does the user come?

<Charlie> in particular why SCXML was needed as a new markup language vs. just using XMI which allows for UML diagram interchange

DSR: Thre is a distinction between interop in authoring tools and what you actually deliver. It will be increasingly complicated. Enabling more people to author more simply and reliably including composability, is important

JOC: There are problems with old browsers (IE...) that don't get anything right.

DSR: Xforms might be just the right level of abstraction to enable authoring tools for people who would run a mile from the makup needed

HH: We found that XHTML and so on is not enough to describe the semantics. The variety of devices is so big that having the "discussion" between the source and destination is important to create a hihgh level declarative apporach to writing content

KS: Tractability becomes traceabiliy - you can follow what is happening all through the system which makes it easier to find and change stuff..

ARIA, accessibility, multiple modalities

JOC: ARIA is good for describing applications, and it is really needed as Web 2.0 applications are developed on HTML etc, lacking the ability to provide something useful to screen readers and other multi-modality scenarios.
... guess that ARIA is the most interesting option...

<mikko_honkala> Is this the correct link? http://www.w3.org/TR/aria-roadmap/

CMN: ARIA is a hack. It just lets you glue on a bit of stuff to fix up badly-authored content. On the other hand, most content is "badly authored", and will be for a long time. ARIA gives people a way that is readily accessible to solve the problems without waiting for them to pick up the new languages and tools that would make the world perfect, so it has an imprtant role to play. Especially in the time it takes for tools to improve and get


JOC: There is a role for ARIA to support people who are using old technologies, and there are a lot of people who are going to fall in the gap it covers.

DSR: Think we can distinguish user tools, author tools, and accessible author tools. It is costly for people using poor tools to get it right.

JOC: There will always be tools taht make dirty markup. A lot of developers don't really know HTML deeply enough to make a really clean tool.

DSR: Should move away from focusing on markup language for authors and look at what ae the tasks that are difficult, where we can provide good ways to increase productivity and quality at reduced costs?
... Which leads to the better tools being able to win out, and gives authors reasons to do stuff better.
... Haven't seen large uptake in multimodal, but accessibiltiy and multi-modality are hooking into the same way

FP: Accesibility benefits from declarative approaches.

Concrete vs Abstract Authoring

DSR: There was work coming out about describing high level stuff like tasks and abstract UI, and otehr stuff that was more UI-centric

FP: We have been working on the distinction...

CW: There is the nice kind fo framework, but nderneath is the issue that the developer has about making that work. Marxists say :do as we want and it will work", Capitalists will say "you have to figure out how to make what people want to do into the results Marxists promise"

Bruce: Tis is like the dclarative and intent-based stuff
... we really want people to be able to express teir intent. If they won't do that, how are we going to find out what they were trying to achieve

DSR: If you can provde people tools that help them externalise what they are trying to do, it makes it easier to get them to give the information. People like WYSIWYG stuff, not markup and properties and attributes.
... WikiMarkup and its ilk are not that effective. Powerpoint allows zillions of people to make (horrible) slides with titles and stuff on them.
... there are abstractions that fail. CSS doesn't make it trivial to do these. Expense reports don't generally include the workflow of approvals and confidentiality in the document.
... If you describe the interoperability requirements end-to-end then you can get better results.

LP: People don't see the value of providing more abstract information, because they cannot really easily produce the end-to-end chain.
... we should model the interaction flow between pages and which jumps are possible now...

LV: If we model on coordination, I would assume there is more interest in adding the information because it can be directly used in a way that people see the value.

KH: What do you mean by adding?

LP: If you have a website, there isn't a summary that describes the structure of the site. If you crawl all the links you would have to be filling in forms and so on that makes crawling ineffective.
... having processs logic information that could show a value, people would use it.

RL: Hope we will talk about the things that will let us do that.
... We are on a journey. This WG is trying to understand which steps can be taken now, and which need more work before we can get them into the universe
... you need to analyse what authors are trying to achieve, but that isn't rocket science - it has happened before.

CW: People won't do work without obvious value. For us, it is about adding composability and service composition. The benefit is clear, and that is waht motivates people to factor the stuff, and that makes the pageflow information emerge from the content.

JOC: If a screen reader user wants to browse a page, they could extract the headings or links, and follow that. Within that modality, that little section is effectively a page to that user. There is huge potential for expanding the metadata associated with those links.

LP: Yes. If you can get some more of that data you can find out things like where you are going to go in an interaction...
... if that could be revealed, it would improve the way you could devliver stuff

[/me thinks of one of the early presentations]

Rhys Lewis wants to agree with Chaals

RL: It is true that right now people don't have tools that do everything nicely. The goal is to make the good tools mainstream for the community - to be the things they get stuck in their shopping bag and use...

SS: I am looking at end-to-end development, and would view these concepts belonging to 3 groups - user interfaces, dialogue flows, and abstract data models.
... linking thes together could be helpful
... I am trying to get concepts for the M, V and C of MVC. If there are good enough tools, we could pass the problem back to the renderer.

DSR: Looking at tools for ordinary people, you need to provide appropriate stuff that supports the tasks people are trying to achieve. There are people using tools, and at another end you have people building libraries and components, with programmers in the middle. Building up all these layers is important, and the second-rate tools will fade out if the competition is better.

Functional User Experience

KS: You need to get something that works, even if it ain't pretty.

RL: We tried to decide what we meant by "works" in DI. We came up with "if the user can achieve the objectives the author had in mind, then it works". The other end was a harmonised experience, that was also nice to use (at least in the way the author conceived of that).

SS: Would it be funcitonal if the user could get what the author wanted but not what the user wanted?

RL: We fell on the author because they are the one who has an expectation that should be matched by the process of building it. You could find something that was good for the user, not the author, and that's all very well.

RH: Thatis why Web 2 is working - you can mashup stuff that wasn't considered before

FP: I am wondering if the terminology is right... how functional is functional...

JOC: The definition of usability is often task oriented, whereas quality of experience matters to users.

CMN: Right - accessibility relies on actually functional experiences, and where you don't look at how pleasant/easy/fun it is then you are not doing a very good analysis in terms of telling peope what options are good or not

JOC: Yeah, quality is really important.
... accessibility means usability analysis needs to be done in different ways and the results might come out very differently

LP: If we put the bar too low for things on a technical level it may not be feasible to make things usable in a nice way. If you require content authors to support a lot of lower level, they forget to do the hgher-level things. When I am on a mobile, people have done some optimisation that excessively "optimises" to give me something that is worse than I want.
... we should aim high...

DSR: Minimally functionl reduces the number of people who can really use it.

SS: There are subjective and objective criteria - the task oriented question of efficiency, and how much fun people have doing it
... where do we put the balance between these two aspects?

LP: Steven started with the computation power theme. If we start standardisation and finish in a year, whenthe world catches up, a couple of years have passed. No point aiming it at the stuff that is around now, but a couple of magnitudes more power around...

CMN: The simple basic machines are still around and still being made new, so you need to work out how to match a bigger range than before, not ust shift the goalposts. We are seeing more devices now that don't match the high-end specs of 1992 than there were in 1992. So we need to do more than just aim at one particular point.

SS: [editor missed]

MH: Today the inteesting applications are spreadsheets, maps, ... Maybe when this stuff comes out the interesting stuff is maybe 3D vector graphics...

DSR: Some of the devices being brought in have very limited processing power, and we want to see how to make it easy to author applications for those.

meeting adjourned

<Rotan> http://www.dublinpubscene.com/thepubs/breakfortheborder_map.html

<Steven> presetn+Marie-ClaireForgue

[End of minutes]