Hypertext Coordination Group Backplane Meeting, Amsterdam, NL
20 Nov 2006


See also: IRC log


John_Boyer, [IPcaller], Marcos, +31.20.592.aaaa, Steven, Charlie, Rhys, Dave, Rafah, Dick, Jack, Chris, Scott, Jose_Cantera, Mikko_Pohja, ht
Mark_Birbeck, Kevin_Kelly
Steven and Charlie




<Steven> is IPCaller Marcos?

<Marcos> yep, thats me

<Steven> thanks

<Marcos> thanks!

<Steven> Like to identify yourself for the record?

<Marcos> Marcos Caceres, with the WAF WG (but not representing them formally), from Queensland University of Technology, Brisbane Australia :D

<Steven> Welcome

<Steven> So it's 9pm where you are?

<Marcos> 6pm

<Marcos> close :)

scribenick Dave

<scribe> ScribeNick: Dave

Steven introduces the meeting and comments that last two years have seen the warmest autumn here in Amsterdam since records began.

Sinter Klaas (Santa Claus) visited the city yesterday showing children with candy.

This is the dutch equivalent of Father Christmas and Steven notes that these are two different people.

Father Christmas gained his red robes after a marketing campaign by Coca Coala!

<ChrisL> the joys of multiple inheritance

Steven introduces the backplane (slide 3)

<ChrisL> http://www.thehollandring.com/sinterklaas.shtml

<ChrisL> Minutes will be public. Presentations too (subject to approval by each speaker)

Charlie is standing in for Kevin who was unable to attend.

Chris notes that the Backplane Note was published last week, see http://www.w3.org/TR/backplane

We go around the participants introducing our names and affiliation.

<ChrisL> Backplane note was published last week

<ChrisL> http://www.w3.org/TR/2006/NOTE-backplane-20061116/

<jcantera> hi

<jcantera> has the meeting started?

Charlie starts with the backplane presentation review, see http://www.w3.org/MarkUp/CoordGroup/2006/wiecha.pdf

Chris notes that Charlie has omitted XML as a platform technlogy, but this is probably taken for granted.

Charlie says yes.

The stuff on the bottom (platform technologies) we take for granted.

I hope that RDF joins it soon.

The stuff on the top is our target.

<ChrisL> CW: Important to define the mapping between declarative and DOM-based approaches (eg declatrative events vs DOM event handlers)

Charlie would like to see future WG agendas mention what it is each WG is making use of and contributing to the backplane (the middle layer in his diagram)

He talks us through an example involving SCXML.

Chris: can you clarify the distinction between hookup and mashup?

Charlie: hookups involve some glue in the middle that is needed to join the components.

The example makes use of XBL from the WAF WG for binding.

SCXML is running in Mozilla, and used to monitor events and to drive behavior.

The other components have no a priori knowledge of each other.

Steven mentions Mark B's work on using XBL, XForms and Google Maps.

<ChrisL> could this scxml-based example be mounted on the w3c site so people can examine it?

Charlie's map demo uses an XForms data model for coordinatin between the XForms Model and the Yahoo! map widget.

<Steven> Charlie types in a zipcode, and on tabbing out the map is updated

<Steven> then types a search for sushi

<Steven> then flags come up showing locations for sushi

<Steven> clicking on the flags sends messages via SCXML back to the XForms model

<ChrisL> event-based control logic

<ChrisL> component can be reused without rewriting for different applications

Steven: I've analysed Mark's mashup vs Google's own code. Mark's is very much smaller.

<ChrisL> declarative and procedural (dojo was mentioned) components can work together

<Steven> 200+k vs 25k lines

We now move to John Boyer's slides, see http://www.w3.org/2006/03/01-Boyer/IntelligentXML_Boyer.1.html

Intelligent XML Data Model.

John suggests that compound documents based upon the backplane can play a role for some of the server-side functions hitherto.

This architecture is orthogonal to whether you choose to use a workflow approach or not.

The diagram indicates that software package cost is inversely related to web application development complexity.

In other words, we can reduce the complexity of developing web apps through the use of declarative approaches together with the associated functional components.

Ajax is very handy as a low level mechanism.

He likens it to the assembly language of the Web.

He moves to the last slide (the pyramid diagram).

John describes his notion when you delete a node in the data, not only do its children (XML content) get deleted, but also all of the dependent views etc. get cleaned up too.

He wants to see a submission mechanism that works for both XForms and other kinds of data model.

Steven adds that the separation between data and presentation is crucial to adapting the view according to the device capabilities and choice of modalities.

<Steven> No, the separation of 'composition of view data' from the model data, and actual presentation on a device

Charlie suggest that John adds in examples of controllers such as SMIL and SCXML.

(to the pyramid diagram)

Essentially replace "sig*" by "SMIL,SCXML"

John explains that "Sig*" refers to signature generation.

We take a 10 minute break and consume special Sinter Klaus biscuits (Speculaas Molens)

<John_Boyer> Alas, my phone needs to sleep (and me too). I hope the rest of your meeting goes well today.

sleep well!

<John_Boyer> Thanks Dave

<ht> HST observes that Boyer's use of signature presupposes that we rewrite XML Signature to work on infosets/data models, which would in fact be a good thing

<ht> HST asks if the observation that there was conversion on XPath is consistent with CSS

<ht> CL and SP point out the CSS Selectors is at least apparently a competitor

<ht> SP discusses the red herring of incremental display

<ht> CL (and HST) note that namespace-awareness is an important aspect of lower-case 's' selectors

We resume after a lively discussion in the break.

Steven presents "events", see http://www.w3.org/2006/03/01-steven-tp-events/

Steven argues for a consistent approach to events across W3C.

He would like to move away from device specific event names to device independent names. An example is DOMActivate in place of click.

<Mitsub> HT: Need for xml pipeline to deal with event handling

<RhysL> +1 for the ability to separate binding of events to handlers from markup

Chris: you need to deal with precedence between declarative and scripted event handlers.

Dave notes that latency issues in distributed applications may preclude pipelining in some cases.

Further discussion is needed on security issues when binding across documents.

<Mitsub> late binding of intentional events to actual hardware

i.e. the means to hook into the intentions without needing to know about the implementation details.

Chris asks about updating the backplane document to refer to DOM3 events rather than DOM2.

Steven says there is a new version in preparation, but it isn't yet public.

The published backplane note also includes lots of links to W3C Member only materials.

XML Events version 2 will support DOM3 events, adds Steven.

<Mitsub> Temporal scope of events, SMIL events revision

Dick says that SMIL WG is very interested in temporal issues for events.

Rhys asks about the status of the XML Events spec?

Steven responds a new members only draft is expected soon.

<scribe> ACTION: Steven to consult editor of XML Event spec and report back to HCG. [recorded in http://www.w3.org/2006/11/20-backplane-minutes.html#action01]

We have a quick summary of the changes introduced in DOM3 events.

<scribe> ACTION: Charlie to update backplane note to remove non-public links. [recorded in http://www.w3.org/2006/11/20-backplane-minutes.html#action02]

Steven's link to a brainstorming session on intent based events in the agenda is something we will come back to when we discuss next steps for the backplane work.

<Marcos> URI?

He moves on to Mark B's presentation on submission (Mark was unable to attend the meeting)

<MikkoPohja> http://www.w3.org/MarkUp/CoordGroup/2006/birbeck.pdf

<Marcos> thanks

Mark's slides describe Ajax and then the XForms submission element.

What is needed to bind Ajax to a submission element?

Mark notes that XMLHttpRequest is too low level to represent what is needed.

We need a better mapping to a DOM API. Some of the best applications involving Ajax make use of wrappers such as Dojo and TIBET, etc.

In particular we need more notification events and support for synchronization.

e.g. TIBET allows you to do things when two submissions are done.

(analogous to SCXML join)

Rafah: Is the submission module being worked on within the XForms WG?

Steven: yes.

We now switch to Rhys's presentation, see http://www.w3.org/2006/Talks/1120-rl-wai-di/

Al Gilman is unfortunately unable to be present, and Rhys is stepping in on Al's behalf.

<Steven> welcome back Jose

<jcantera> ok

<Steven> Would you like to introduce yourself (on irc) for the record? (WGs, location, time zone, employer?)

<Mitsub> gestures and actions - should be defined by the user, not by individual pages

<jcantera> yes, sure

Rhys refers back to the common user interface guidelines for OS2.

There is prior work on standardizing actions in specific domains, and providing bindings to gestures for invoking them, e.g. through spoken commands.

Gestures can be used for both input and output.

Gestural output may involve animations.

Rhys argues that the bindings between gestures and actions belong to the delivery context, which includes user preferences, device capabilities and environmental conditions.

Rafah explains that the DCI may be used server-side as well as device-side.

DCI = delivery context interfaces, see http://www.w3.org/TR/DPF/

Charlie notes that we need a well defined binding mechanism, and that this is separate from the issue of whether the actions are concrete or abstract.

Henry: we probably don't want the same binding mechanisms for UI actions and application level events

He believes that there is a distinction between UI events and events such as submit this page.

<ht> HST expresses his unclear concern that the level of actions/gestures/binding which Rhys has just talked about is not just different in level, but qualitatively different from the kind of binding which is at the heart of the management of control flow for the backplane

Dave notes that for application level behavior semantic events are valuable and there is an interpretation layer between them and UI specific events such as you get with XHTML.

The semantic interpretation layer is application dependent.

Dick asks if this involves using RDF. The answer is that it isn't necessary to involve RDF.

Rhys draws attention to the privacy issues involved in disclosure of personal preferences.

<ht> HST trys again "Using events/binding for Loose Coupling of Applications feels very different to me from the kind of binding that goes on between gestures and actions"

We now move to Rafah's presentation, http://www.w3.org/MarkUp/CoordGroup/2006/wiecha.pdf

Rafah is interested in the loose coupling of web application components.

She talks us through the demo involving a combination of XForms and a Yahoo! Maps widget.

The combined widget just looks like an XForms composite widget.

Components can be combined in a nested fashion arbitrarily deep.

The backplane provides a means to achieve this.

She shows us the markup for the example.

The trick is the means to submit data to the enclosing component.

This isn't yet part of the XForms submission framework.

Dave notes that this is very similar to the MMI WG's view of modality components which are black boxes that may be composed of subsidary components in a russian doll like fashion. These may have their own data model which is hidden from the outside.

Rafah says that in her work, the components may expose a data model and bind that into the parent component.

She talks about the idea of distinguishing between public and private data models.

<Charlie> ping

VBWG contributions to backplane

<ChrisL> Scribe: ChrisL

Data model requirements, eventing (in SCXML and VoiceXML), and data submission (strategies/requirements from SCXML and VoiceXML). (Scott McGlashan, Rafah Hosn)

slides at Data model requirements, eventing (in SCXML and VoiceXML), and data submission (strategies/requirements from SCXML and VoiceXML). (Scott McGlashan, Rafah Hosn)

slides at http://www.w3.org/MarkUp/CoordGroup/2006/VBWG-backplane-04.pdf

Presentation layer in data flow presentation (DFP) model does not require speech output.

data abstraction/data hiding in components

the different languages have different data models, but all exhibit containment

SP: Similar to JSON

application scope allows data sharing between documents

<RhysL> JSON is a streaming form of representation for ECMAScript objects. I think Scott means the objects within the ECMAScript associated with VoiceXML

submission is currently string based, structure is not preserved.

voicexml does not require a DOM implementation

however, fits better with DOM3 rather than DOM2

scribe: event categories

scxml 1.0 is uml 2.0 compliasnt but extends constraints to ensure it is deterministic

can combine events with conditions (event plus condition moves to new state)

HT: can mint URIs for any state? these are unique IDs? (yes)

LK: How do you make these pluggable? (all platforms support both ecmascript and xpath expressions)

(dicsussion on determining what syntax is being used in a given instance)


there is no way. type attribute on root element seems to be a popular suggestion

w3c xml schema is likely choice for validation

LK; what id the schema location evaluates to a nodeset (error)

scribe: better to take the first node in the set

<klotz> Please examine a rule for converting the assign/@location nodeset to a document for schema validation. //@* isn't a good idea. //foo could also result in a nodeset and you may wish to specify //foo[1].

invoke allows spawning of separate processes

<klotz> please read xsi:schemaLocation, xsi:noNamespaceSchemaLocation and xforms:model/@schema to look for subtleties and limitations in scxml:assign/@schema

SP: surprisingly large overlap with XForms

JJ: Which bits of XForms were *not* copied?

mainly the use of exmascript expressions as well as xpath expressions depending on which is more convenient

RL: diselect is pluggable, so host language could specify which expression language is used. sam for scxml, could depend on the host language

LK: The integration should say what the expression language is

CW: Should be everywhere, odd to have some attr flexible and some not

DB: Choice can be influenced by what other components are being integrated and what they use.
... prefer to reference rather than reinvent

s/my SIM/his SIM/

<dino> I was hoping to "arrive" earlier but I only just got home from travel

LK: XForms does type validation on each update, but structur evalidation only just before submission
... can have invalid data; instance is typically invalid before user enters required data

<scribe> Agenda: http://www.w3.org/MarkUp/CoordGroup/2006/backplane-meeting.html

JJ: Want to use the existing SMIL system tests

LK: So, reusing constrainsty from one model in another one (yes)

model can be in multiple namespaces and can be validated to multiple schemata

LK: Data can be copied between instances, but is not copied between models (scribe does not understand the difference)

<Steven> A model represents a cohesive set of data, and contains different instances

<Steven> So you might have one model for login and one model for search

<Steven> they are independent


LK: XPath predicate calculations can be expensive. Avoid re-evaluating them constantly.
... decorating the tree with types is much cheaper than re-evaluating the nodes

SM: Concerns on validation at all, in high transaction scenarios


Challenges and Opportunities for the Backplane

Henry S. Thompson


(brief discussion of mixedUIXMLNamespace-33 and ineffability)

CW: real pitfall to say its another 'declarative is good' campaign.

HT mentions

XProc: An XML Pipeline Language

W3C Working Draft 17 November 2006


CW: The blackboard analogy does not apply here
... mechanism vs policy
... and the actors are human (component assemblers)

(discussion on the html browser security model of same-site constraint vs web-wide linking)

(discussion of versioning of components)

DSR: ECMAScript v4 tackled some of that

HT: Putting datamodel at the heart is a good idea
... migration over time from web delivered to substrate-provided functionality

CL: And without having to edit the content - don't expose mechanism in content

saving state.

SP: Bookmarking and saving state not identical

<Steven> Good nmorning John

<John_Boyer> :)

take-home message is to be aware of these tag issues and to document the chosen trade-offs

CL: mentions CDF discussion re granularity of namespace dispatching and separating dispatching from rendering

see http://lists.w3.org/Archives/Member/member-cdf/2006Nov/att-0076/cdf-ns-dispatch.txt

and http://lib.tkk.fi/Diss/2003/isbn9512268043/article8.pdf


<John_Boyer> Backplane is like plugboard for components to interoperate; one component could be something like super SCXML which could indicate the application states where saving state would be a good idea

<John_Boyer> So when you hit the back button, you go back to the last transactional state

<Steven> http://www.chiark.greenend.org.uk/~marisal/ie/ingv.html

MMI input

MMI input for Backplane Workshop

Dave Raggett, <dsr@w3.org>



(discussion on russian-doll nesting, as with previous talks)

fixed(?) set of life-cyle events, in pairs (request and response)

RH: Does the model change in response to these events? (no)

components can be declarative or could be (eg) in Java

CW: lingua franca vs arbitrary extensible events
... need to not get pigeonholed as a markup-only stance

<steve> hi folks

<steve> have a good meeting

modality components can submit without going through a central submit component

Does Backplane centralize submission?

CW: Whats the communication between a modality component and its parent? (delivery context may have info about formats required)


UWA input for Backplane Workshop

Dave Raggett, <dsr@w3.org>


applications are no longer just clients - they become servers as well

device discovery and integration

declarative solutions can be based on proprietary solutions, or on scxml.

JJ: delivery context - mobile presentations a problem or not (not - early and late adaptation both used)
... looks fairly static, rather than adaptation at last possible moment

DSR: its a high level model, showing the range of possibilities - does not indicate the most likely sites of adaptation

OMA DPE is mentioned - more dynamic

(discussion of run-time flow modelling)

DB: How is temporal ordering reflected in the UWA work?

<Steven> V2 universal remote: http://www.imtc.gatech.edu/projects/technology/V2_univeral_remote.html

simple devices that can send and receive events, but have ultra-limited capabilities

CW: XSXML looked at microbroker rfid use of scxml


<rafah> RH: I believe Charlie meant "IBM looked at using SCXML for RFDID inside IBM's micro-broker"

HT: Do these resources have URIs?

DSR: they can do. not always. Whats the URI for a screen size of 120x96?

HT: Just asking to examine how much granularity to expose to Web Arch

proxies and remove eventing - messaging is hidden and might use proprietary protocols

REX - conveys events as an xml payload


extensible as it is not (will not be) a fixed set of events

many different wgs have events they want to transmit remotely

<John_Boyer> Steven, bark

<Steven> woof

DSR: Ajax covers some things but has to poll to get events. REX allows remote listeners

LK: Isn't this an unresolved issue on SOAP, too, for asynchronous calls

OMA DCD dynamic content delivery - 'drip feed'


DSR: worries about phishing have restricted rich solutions because no way to manage privacy and identity
... richer acces to device cababilities considered crucial

<John_Boyer> I have to go now; Have a good evening.

versioning issues on script libraries - but need to cache rather than load multiple times from different sites

privacy concerns over access to browser history, vs need to provide a meaningful 'back' to go back through the states

LK: Undo and redo - how to undo past an action boundary?

RH: Snapshots of data model, with anchor marks authored where they make sense

HT: Seems reasonable to allow that, given that cost of arbitrary state backtracking could be high

LK: Restore to state of the data model - might not be identical to the last time you were there

HT: Loads of database literature on backtracking

need to inform user of any side effects of backtracking

need to author compensation blocks

HT: needs to be up to application developer how much effotrt to put in and how much backtracking to allow.

CW: Tomorrow, we will start with scoping discussions

also transitional strategies for incremental development (no flag day)

scribe: then end with next steps

<John_Boyer> Response to HT: Yes this is what I meant earlier about having a pluggable component that would describe the application states worth saving

RL: Good to identifuy any effects on UWA charter

LK: Expression language plugsability would be good to address early

Summary of Action Items

[NEW] ACTION: Charlie to update backplane note to remove non-public links. [recorded in http://www.w3.org/2006/11/20-backplane-minutes.html#action02]
[NEW] ACTION: Steven to consult editor of XML Event spec and report back to HCG. [recorded in http://www.w3.org/2006/11/20-backplane-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.127 (CVS log)
$Date: 2006/11/20 16:33:01 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.127  of Date: 2005/08/16 15:12:03  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/Sint/Sinter/
Succeeded: s/Klaus/Klaas/
Succeeded: s/9eg/(eg/
Succeeded: s/XForms UI/XForms Model/
Succeeded: s/SO/No/
Succeeded: s/9/(/
FAILED: s/dicus/discus/
Succeeded: s/9/(/
FAILED: s/my SIM/his SIM/
Succeeded: s/soes/does/
Succeeded: s/ECMA/ECMAScript/
Succeeded: s/ned to/need to/
Succeeded: s/recieve/receive/
Found ScribeNick: Dave
Found Scribe: ChrisL
Inferring ScribeNick: ChrisL
ScribeNicks: Dave, ChrisL
Default Present: John_Boyer, [IPcaller], Marcos, +31.20.592.aaaa, Steven, Charlie, Rhys, Dave, Rafah, Dick, Jack, Chris, Scott, Jose_Cantera, Mikko_Pohja, ht
Present: John_Boyer [IPcaller] Marcos +31.20.592.aaaa Steven Charlie Rhys Dave Rafah Dick Jack Chris Scott Jose_Cantera Mikko_Pohja ht
Regrets: Mark_Birbeck Kevin_Kelly
Agenda: http://www.w3.org/MarkUp/CoordGroup/2006/backplane-meeting.html
Got date from IRC log name: 20 Nov 2006
Guessing minutes URL: http://www.w3.org/2006/11/20-backplane-minutes.html
People with action items: charlie steven

[End of scribe.perl diagnostic output]