Ubiquitous Web Applications Working Group Charter

Scope

This section sets out the scope of proposed work items, starting with a summary. The Working Group will try to:

• Complete existing DIWG work items on DIAL, DISelect, DCI and CC/PP
• Define a second version of DIAL that incorporates support for SVG, SMIL and XBL
• Leverage the DCI for work on enabling use of device capabilities, remote eventing and device coordination
• Facilitate device and modality independent authoring with work on intent based events
• Organize W3C Workshops on declarative models of distributed Web applications, including autonomous Web agents that act on behalf of, and are controlled by users

The scope for each of these is explained in the following subsections. Please note that this charter defines a number of concrete short term work items together with longer term objectives, which will be approached through the organization of the appropriate workshops as a basis for setting the direction of future work.

Success Criteria

The success of the UWA WG will be measured in the short term by progress on planned work items, and in the long term by their impact on enabling distributed Web applications and the range of devices involved. Success will enable the Web of everything as illustrated below:

Web of Everything Incorporates a very wide range of devices Physical world of sensors and effectors Access to local and remote capabilities Autonomous agents that act upon behalf of their users

Success in achieving this long term objective can be evaluated in terms of:

• The availability of a framework for authoring distributed Web applications that minimizes the costs for dealing with a wide diversity of devices through the use of declarative models encompassing both clients and servers
• The ability of mainstream Web developers to create distributed applications through the use of markup and scripting, where local and remote services are exposed as DOM objects
• Attention to the trust, identity, security and privacy issues as needed to enable widespread use of such applications

The objectives will only be achieved through cooperation with other Working Groups and not in isolation. Coordination is therefore key to the success of the UWA WG.

Complete existing DIWG work items

The intention is to progress existing work items on DIAL, DISelect, DCI and CC/PP along the W3C Recommendation track.

• DIAL combines a number of existing markup languages (XHTML2, XForms and DISelect) into an authoring language that is designed for content adaptation. DIAL is expected to move to Last Call in early 2007. The first public Working Draft appeared in May 2006, and the DIAL Primer in October 2006.
• DISelect provides authors with the means to select different markup according to the delivery context. DISelect entered Last Call in October 2006 along with the associated specification for the Delivery Context: XPath Access Functions, and the Content Selection Primer.
• The Delivery Context Interfaces (DCI) is a framework for dynamic access to personal preferences, device capabilities and environmental conditions as a hierarchy of DOM nodes. DCI was published as a Candidate Recommendation in October 2006.
• CC/PP defines a means for devices to provide delivery context information to servers as a basis for content selection and adaptation. This work item seeks to update CC/PP for alignment with the latest versions of RDF and the OMA UAProf specification which is itself is based upon CC/PP.

Work on a second version of DIAL with support for SVG, SMIL and XBL

Demand for richer interaction with Web applications can be met through increased use of scripting and through combining XHTML with other markup languages. SVG offers resolution independent graphics and animation, SMIL provides for rich coordination of multimedia, and XBL provides a means to add custom user interface controls defined in terms of SVG, CSS and ECMAScript.

This work item will develop a second version of DIAL that incorporates support for SVG, SMIL and XBL, building upon specifications defined by the SVG, SYMM and WAF Working Groups, and the WICD profiles developed by the CDF Working Group. Native support for SVG and SMIL is increasingly common on mobile devices. Including support for these languages within DIAL will give developers the means to support a wide range of such devices, insulating developers from variations in device capabilities and adherence to standards (such variations are taken care of as part of the content adaptation process).

With the demand for richer interaction with Web applications, XBL provides DIAL with a device independent means to bind to rich Web controls. The process of content adaptation can then select the appropriate implementations of these controls to match the delivery context.

A related work item on policy-based layout may be started after the Workshop on declarative models of Web applications. This would build upon the DIWG's advanced layout work item, and preliminary discussions together with the CSS Working Group.

Leverage the DCI for work on enabling use of device capabilities, remote eventing and device coordination

The aim of DCI is to enable Web applications to dynamically adapt to user preferences, device capabilities and environmental conditions. The DCI specification defines platform and language neutral interfaces for a hierarchy of properties expressed in terms of DOM nodes. These properties can be static or changeable, for example, screen size and orientation, battery level, wireless connectivity and field strength. Properties may be both readable and writable, for example, the audio output level. Properties may raise events to notify changes, for example, low battery alerts and loss of network connectivity. The DCI can also be used to expose remote properties, for example, the geographic location of a mobile device, where this is provided by a network-based service.

The DCI defines a framework and not specific properties. Work is now needed on defining an ontology of such properties in cooperation with other groups, both internal and external to the W3C. The resources exposed as DCI properties may themselves change over time as they are bound or unbound. See the following sub-section on device coordination for more details. The DCI may be implemented natively by the markup language interpreter or by extensions implemented as plugins. Note that not all devices will host a Web browser, and may instead support a markup interpreter for an event-driven state machine such as SCXML in combination with the DCI, which provides the means to control other devices via an exchange of events. These include simple sensors and effectors whose behavior is defined in firmware, for example, a remotely controllable air conditioning unit. Event interchange may be based upon REX or other means as appropriate to the device and networking technology.

The DCI can provide a means to respond to external requests. For this purpose, a "server" resource is bound into the DCI that raises an event whenever a request is received. This is useful for agent like applications. The DCI can also be used in situations where a subscribe/notify mechanism is needed. An example, is where the DCI is used to bind to a remote printer. The implementation of the corresponding DCI node subscribes to events from the printer, such as job completed, or out of paper. The messaging mechanisms involved are hidden from the user of the DCI.

Enabling use of Device Capabilities

To enable new classes of Web applications, the UWA WG will consider the development of platform and language neutral interfaces for a range of device capabilities. For instance, there is interest in providing mobile applications with safe and secure access to capabilities such as: local storage, audio settings, imaging sensors, and device based applications such as the address book and calendar. These would be exposed via the DCI using the resource binding mechanisms described below. There is also interest in Web page interfaces for controlling the presentation of streaming media, e.g. for use in applications involving home media appliances. The work will be done in cooperation with other groups, both internal and external to the W3C, for example, the W3C Web API Working Group, the OMA for requirements relating to applications involving mobile devices, and the DLNA and the Peer-to-Peer Universal Computing Consortium (PUCC) for applications involving network appliances.

Remote Eventing

Web page scripts are event driven. The UWA WG seeks to extend this to allow Web applications to respond to externally sourced events delivered via the DCI, and to target events at DCI-based DOM nodes that act as proxies for remote services. Remote events for XML (REX) is an extensible XML grammar for events that forms a basis for passing events between components in a distributed Web application. REX can be used to support remote event listeners and remote event dispatch. This can be applied, amongst other things, to provide remote user interfaces to devices, for example, enabling the use of a mobile phone for controlling the air conditioning and lighting in a meeting room. This work started as a joint product of the Web API and SVG Working Groups. The UWA WG will complete this work and extend it to cover a wide range of DOM3 events and other event sets. Additional work is anticipated on mechanisms to support asynchronous dispatch of remotely sourced events delivered via the DCI, and bindings to suitable transport mechanisms in cooperation with other groups.

Device Coordination

Ubiquitous Web applications require access to a wide variety of resources, both local and remote. To simplify authoring, such resources are exposed as objects in the document object model and accessible from web-page scripts. The challenge is to provide Web authors with a simple means to identify the resource they want, and to bind it into the session as a DOM node within the DCI framework. Once bound the node can raise events and be the target of events. The author is thereby shielded from the details of how the resource is realized and the communication mechanisms used to support it. The following diagram indicates how DOM objects can act as proxies for local or remote services as part of distributed Web applications.

Resources can be identified by naming them or by providing a description, either directly, or indirectly by a link to the description. A platform and language neutral interface will be defined for binding resources and returning the corresponding DOM object. The binding request may involve a reference to a named binding service. This enables the use of Web servers as part of the binding process. The binding mechanism may also be layered on top of existing device coordination services. When resources are bound, this is signalled by a DOM event, and similarly, another event will be raised when the resource is unbound.

Resource binding takes place in the context of a session, which itself is a resource. Sessions may be temporary, for example, only lasting for the lifetime of a Web page, or permanent, such as for meeting rooms, which applications can join and leave. Attention will be given to the trust, identity, privacy and security issues associated with access to resources, in cooperation with other groups, starting with the workshop planned for early 2007.

Facilitate device and modality independent authoring with work on intent based events

User interfaces for Web applications need to provide users with information about the current context such as where am I, what is there and what can I do? To enable accessibility interfaces and to support device independence, it is valuable to be able to describe this in terms that abstract away from particular user interface gestures that depend on the device capabilities and the modes of interaction available to the user. A work item on intent-based events is proposed to define an ontology of actions and associated events. This has been initiated as an outcome of W3C Backplane Workshop held in November 2006. It is anticipated that this work will be carried forward in a joint task force involving several W3C Working Groups, with the Ubiquitous Web Applications Working Group having the responsibility for driving the work along the W3C Recommendations Track.

W3C Workshops on declarative models of distributed Web applications

The Working Group plans to organize two W3C Workshops as a basis for guiding future W3C work:

• June 2007 - Workshop on Declarative Models of Distributed Web Applications. This will be co-located with the first face to face meeting under this Charter.
• Spring 2008 - Workshop on Autonomous Web Agents that act on behalf of, and are controlled by their users. Location and date to be decided.

The following provides a tentative look at the subject areas for discussion at these Workshops. The scope would be refined by the programme committees formed for each Workshop:

Declarative Models of Distributed Web Applications

This Workshop is scheduled to take place in Dublin, Ireland on 5 -6 June 2007, hosted by MobileAware with the support of the Irish State Development Agency, Enterprise Ireland, and co-chaired by Dave Raggett (Volantis Systems) and Kevin Smith (Vodafone).

The variations in capabilities and conformance to standards across devices pose considerable challenges to application developers. Declarative models of Web applications can support device and modality independent representations that reduce the cost of developing individual applications. The adaptation process can be customized via policies that select the content, presentation and behavior appropriate to specific delivery contexts. Preliminary requirements discussions on device-aware layout have already taken place between the Device Independence and CSS Working Groups as part of the DIWG's Advanced Layout work item, but there are opportunities for broader kinds of policies, and an increasing role for protocol independent metadata. for devices and services as a basis for enhanced interoperablity.

DIAL is part of the tradition of model-based user interface design languages. There are opportunities to combine model-based approaches with other techniques, such as dialogue-based formalisms. Such approaches can be applied to describing Ubiquitous Web applications as a whole rather than the portions that run on particular devices. There is considerable potential for advanced content management tools.

Another consideration for the Workshop are the challenges raised in dealing with trust, identity, privacy and security. The current sandbox model for browsers involves the same site restriction on sending HTTP requests via XMLHttpRequest. This is inadequate for applications involving multiple devices, e.g. a browser, a remote printer and a web server, where you want to be able to exchange events with the printer. The sandbox is also a problem for richer access to device capabilities - something that is critical for Ubiquitous Web applications. One approach is to ask the user whenever the Web page wants to break out of the sandbox. This gets tedious upon repetition, and permission may be needed in advance for applications that are run in response to events, when the user may not be around to ask. Such markup may, for example, be run on a Web server rather than on a Web browser.

To avoid asking the user each time, there needs to be some means to check that the application hasn't been tampered with. This runs into problems with versioning and with pages with customized content such as the news headlines or the latest weather. Another approach is when the page is signed by an organization that the user trusts. The user can then be asked whether she always trusts applications signed by that organization. This tends to raise additional challenges, such as applying the notion of identity to Web pages, to groups of pages acting as a Web application, to devices, and to users and groups of users. A distributed framework for making assertions and making inferences on them would be valuable, and should cover trust, identity, privacy and security. Users should be aware of what they are being asked to commit to, for example, what device capabilities or sensitive information the application is seeking access to. This in itself presents significant usability challenges to avoid overloading the user, with the risk of inappropriate decisions.

There are also major usability concerns with asking the user to make trust related decisions for devices with small screens for which the entire display may be under application control and therefore at risk from spoofing attacks. One way around this is to utilize authentication mechanisms that renders spoofing attacks unprofitable, for example, by exploiting SIM-based security services that rely on the physical presence of the SIM card. Another concern is how to associate devices when they are introduced into home or office environments. For example, when using a wireless connection for printing, the printer with the strongest signal may be in an adjoining apartment, or in the office of the company located on the floor above. The usability issues for managing home networks are particularly challenging and this impacts upon security.

Autonomous Web Agents

With the ability to respond to external events, and to access local and remote resources, Web pages are no longer acting merely as clients in the client-server model, but take on the role of agents. Such agents can be driven directly by end-users, or indirectly as autonomous Web Agents (or wbots) that act on behalf of, and are controlled by users. Such agents can be represented as a combination of markup and scripts. Markup languages for event-driven state machines such as SCXML (based on UML's Harel state tables) look especially suitable for use in describing agent behavior, leveraging widespread experience with UML.

Agents will be useful on a wide range of devices, including mobile devices and big websites, and can act on behalf of users at all times of the day or night, whether or not the user is working, relaxing, asleep or busy, or out of range of network communications. Agents are also expected to play a role in coordinating distributed Web applications, for example, an interaction manager that runs on a network server and which coordinates remote user interface components. Agents may be situated on large websites or on devices hidden behind firewalls or NAT. In such cases, agents accessible via public URIs (e.g. on public websites) can act as proxies for agents that are either hidden or turned off or otherwise unreachable.