This document outlines the requirements for a CC/PP framework, vocabulary, and trust model, and provides an overview of an architecture that satisfies these requirements. It represents the current consensus of the working group.
This document is a working draft made available by the World Wide Web Consortium (W3C) for discussion only. This indicates no endorsement of its content. This is work in progress, representing the current consensus of the working group, and future updates and changes are likely.
The working group is part of the W3C Mobile Access activity. Continued status of the work is reported on the CC/PP Working Group Home Page (Member-only link).
It incorporates suggestions resulting from reviews and active participation by members of the IETF CONNEG working group and the WAP Forum UAprof drafting committee.
Please send comments and feedback to email@example.com.
A list of current W3C Recommendations and other technical documents can be found at http://www.w3.org/TR/.
The goal of the CC/PP framework is to specify how client devices express their capabilities and preferences (the user agent profile) to the server that originates content (the origin server). The origin server uses the "user agent profile" to produce and deliver content appropriate to the client device. In addition to computer-based client devices, particular attention is being paid to other kinds of devices such as mobile phones.
The requirements on the framework emphasize three aspects: Flexibility, extensibility, and distribution. The framework must be flexible, since we can not today predict all the different types of devices that will be used in the future, or the ways those devices will be used. It must be extensible for the same reasons: It should not be hard to add and test new descriptions. And it must be distributed, since relying on a central registry might make it inflexible.
The basic problem that the CC/PP framework addresses is to create a structured and universal format for how a client device tells an origin server about its user agent profile. This work aims to present a container that can be used to convey the profile, and is independent on the protocols used to transport it. It does not present mechanisms or protocols to facilitate the transmission of the profile.
The framework describes a standardized set of CC/PP attributes - a vocabulary - that can be used to express a user agent profile in terms of capabilities, and the users preferences for the use of these capabilities. This is implemented using the XML [XML] application RDF [RDF]. This enables the framework to be flexible, extensible, and decentralized, thus fulfilling the requirements.
RDF is used to express the client device's user agent profile. The client device may be a workstation, personal computer, mobile terminal, or set-top box.
When used in a request-response protocol like HTTP, the user agent profile is sent to the origin server which, subsequently, produces content that satisfies the constraints and preferences expressed in the user agent profile. The CC/PP framework may be used to convey to the client device what variations in the requested content are available from the origin server.
Fundamentally, the CC/PP framework starts with RDF and then overlays a CC/PP-defined set of semantics that describe profiles.
The CC/PP framework does not specify whether the client device or the origin server initiates this exchange of profiles. What the CC/PP framework does specify is the RDF usage and associated semantics that should be applied to all profiles that are being exchanged.
In this section, we describe use cases for the use of a profile to adapt information for presentation by a client. The use cases are intended to describe a number of situations where the CC/PP format can be used; however, they are not an exhaustive list, nor are the behaviour of any element in the use cases described below mandatory. Rather, they represent the working groups best understanding at the time.
Using the World Wide Web with content negotiation as it is designed today enables the selection of a variant of a document. Using an extended capabilities description, an optimized presentation can be produced. This can take place by selecting a style sheet that is transmitted to the client, or by selecting a style sheet that is used for transformations. It can also take place through the generation of content, or transformation, e.g. using XSLT [XSLT].
The CC/PP Exchange Protocol [CCPPex] extends this model by allowing for the transmission and caching of profiles, and the handling of profile differences.
This use case in itself consists of two different use cases: The origin server receives the CC/PP profile directly from the client; the origin server retrieves the CC/PP profile from an intermediate repository.
This use case describes a case where the profile is used by an origin server on the web to adapt the information returned in the request.
The CC/PP Exchange Protocol [HTTPex] creates a modified HTTP GET which carries the profile or the profile difference. The extended content negotiation with the CC/PP Exchange Protocol uses the CC/PP format to describe the device. In this context, no vocabulary has been defined. The existence of the CC/PP Exchange protocol is assumed.
When the interaction passes directly between a client and a server, the process is relatively simple: The user agent sends the profile information with the request, and the server returns adapted information. The interaction takes place over an extended HTTP method, as described in the CC/PP Exchange framework.
When the profile is composed by resolving inline references from a repository for the profile information, the process is slightly more complicated, but in principle the same.
The interactions take place in more steps, but is basically the same:
1: Request from client with profile information.
2: Server resolves and retrieves profile (from CC/PP repository on the network), and uses it to adapt the content.
4: Server returns adapted content.
5: Proxy forwards response to client.
Note that the notion of a proxy resolving the information and retrieving it from a repository might assume the use of an XML processor and encoding of the profile in XML.
In case the document contains a profile, the above could still apply. However, there will be some interactions inside the server, as the client profile information needs to be matched with the document profile. The interactions in the server are not defined. This group will not standardize the matching algorithm. We will standardize the profile system, which is the interface to it.
No formal requirements document has been formulated in this activity. However, the use of the CC/PP Exchange Protocol is assumed.
One requirement is that the integrity of the information is guaranteed during transit.
In the proxy use case, a requirement is the existence of a method to resolve references in the proxy. This might presume the use of an XML processor and XML encoding.
The privacy of the user needs to be safeguarded. This is not a technical requirement, but a societal.
The document profile and the device profile can use a common vocabulary for common features. They can also use compatible feature constraining forms so that it is possible to match a document profile against a receiver profile and determine compatibility. If not, a mapping needs to be provided for the matching to take place.
This use case describes a mixed architecture with an active proxy in the network. The use of CC/PP profiles in the WAP architecture is more fully described in [UAProf].
In the WAP Forum, an architecture optimized for connecting devices working with very low bandwidths and small message sizes has been developed. In version 1.2 of this architecture, a format based on RDF is used to communicate client profile information.
The WAP Forum architecture is based on a proxy server, which acts as a gateway to the optimized protocol stack for the mobile environment. It is to this proxy which the mobile device connects. On the wireless side of the communication, it uses an optimized, stateful protocol (Wireless Session Protocol, WSP [WSP]; and an optimized transmission protocol, Wireless Transaction Protocol, WTP [WTP]); on the fixed side of the connection, it uses HTTP [HTTP/1.1]. The content is marked up in WML, the Wireless Markup Language of the WAP Forum.
The mobile environment requires small messages and has a much narrower bandwidth than fixed environments.
When a user agent profile is used with a WAP device, it looks like this:
The user agent profile is transmitted as a parameter of the WSP session to the WAP gateway and cached; it is then transferred over HTTP using the CC/PP Exchange Protocol [CCPPex], which is an application of the HTTP Extension Framework[HTTPex].
Note that the WAP system uses WML[WML] as its content format, not HTML. This is an XML application, and the adaption could for instance be transformation from another XML format into WML.
The WAP UAPROF group has developed a core vocabulary for mobile devices, as presented in [UAProf].
The WAP Forum went through an extensive requirements gathering process before producing their specifications. The requirements document lists a number of requirements that has been placed on the solution. These have been included in our requirements with the prefix UAprof. For more information, see [UAProf].
Addressing XML fragments
The Conneg working group in the IETF has developed a form of media feature descriptors, which are registered with IANA. Like the CC/PP format and vocabulary, this is intended to be independent of protocol. The Conneg working group also defined a matching semantics, based on constraints.
The CONNEG framework defines an IANA registry for feature tags, which are used to label media feature values associated with the presentation of data (e.g. display resolution, color capabilities, audio capabilities, etc.). To describe a profile, CONNEG uses predicate expressions ("feature predicates") on collections of media feature values ("feature collection") as an acceptable set of media feature combinations ("feature set"). The same basic framework is applied to describe receiver and sender capabilities and preferences, and also document characteristics. Profile matching is performed by finding the feature set that matches two (or more) profiles. This involves finding the feature predicate that is equivalent to the logical-AND of the predicates being matched. (See RFCs [RFC2506] and [RFC2533] for further information.)
The Conneg system is not dependent on any one protocol, and could conceivably be used in all the use cases above, given that the information set to be matched could be expressed in the Conneg format.
Conneg is protocol independent, but can be used for server-initiated transactions.
Here is what the server-initiated use case would look like:
The Conneg system uses a structured text format to describe the device and its capabilities. This is described in [RFC2533].
The Conneg working group has defined a set of core media features, which are described in [RFC2543]. A comparison of these and the UAPROF features is found in [UAProf].
The Conneg group started its work with developing an extensive requirements document, which is found in [RFC2703]. They are listed as CONNEG in appendix 2.
This use case describes a push situation, where a broadcaster sends out an information set to a device without a backchannel. The server can not get capabilities for all devices. So it broadcasts a minimum set of elements. Or a multipart document, which is then adapted to the optimal presentation for the device.
Television manufacturers desire to turn their appliances into interactive devices. This effort is based on the use of XHTML as language for the content representation, which for instance enables the use of content profiles as seen in 2.2.4.
Unlike the use cases above, a television set does not have a local intelligence of its own, and does not allow for bidirectional communication with the origin server. This architecture also applies to several different device classes, such as pagers, e-mail clients and other similar devices.
It is not the case that they are entirely without interaction, however. In reality, these devices follow a split-client model, where the broadcaster, cable head-end, or similar entity interacts with the origin server, and sends a renderable version of the content to the part of the client which resides at the user site. In this aspect, they resemble the intermediary proxy in 2.5.
There are however also use cases where the entire data set is downloaded into the client, and the optimal rendering is constructed there, for instance in a settop box. This resembles 2.1.4, except that the matching of the document profile with the device capabilities does not take place in the origin server, but in the client. In these cases, the CC/PP client profile will need to be matched against a document profile, representing the authors preferences for the rendering of the document.
The CC/PP profile and the document profile should ideally be in the same format, as described in 2.2.4.
The level of detail in a vocabulary that is intended to enable rendering of information on devices without an intelligence of their own is unclear. See section 6.3 for a further discussion on the vocabulary developed by the group, and possible extensions.
Since this can be seen as a special case of the application of document profiles, no separate requirements have been derived.
This use case describes how an intermediate entity interacts with the profile and the information received in the response.
When a request for content is made by a user agent to an origin server, a CC/PP profile describing the capabilities and preferences is transmitted along with the request. It is quite conceivable that intermediate network elements such as gateways and transcoding proxies that have additional capabilities might be able to translate or adapt the content before rendering it to the device. Such capabilities are not known to the user agents and therefore cannot be included in the original profile. However, these capabilities would need to be conveyed to the origin server or proxy serving/ generating the content. In some instances, the profile information provided by the requesting client device may need to be overridden or augmented.
For instance, consider a user agent that can render gif images. If there is a transcoding proxy in the network that has the ability to convert jpeg images to gif, the proxy can assert this capability to the origin server that may be able to render content only in jpeg. Similarly, a gateway may wish to add information to the profile to override the profile information sent by the requestor, for example, to reflect the policies in place by the network or gateway operator.
CC/PP framework must therefore support the ability for such proxies and gateways to assert their capabilities using the existing vocabulary or extensions thereof. This can be done as amendments or overrides to the profile included in the request. Given the use of XML as our base format, these can be inline references, to be downloaded from a repository as the profile is resolved (see use case 2.2.3).
1. The CC/PP compliant user agent requests content with the profile.
2. The transcoding proxy appends additional capabilities (profile segment), or overrides the default values, and forwards the profile to the network.
3. The origin server constructs the profile and generates adapted content.
4. The transcoding proxy transcodes the content received based on its abilities, and forwards the resulting customized content to the device for rendering.
The CC/PP framework can still be used, given the ability of overrides or amendments. It is also possible that users wish to select attributes that can not be overridden.
No specific vocabulary is necessary to express this use case.
This use case requires that entries can be added to the CC/PP profile.
An implicit requirement is that this takes place in a secure way, safeguarding the integrity of the original profile.
The following assumptions have been identified during the process of the work. These assumptions are basic tenets of the design, as well as assumptions about other elements in the process, such as network elements and origin servers. There are implicit requirements inherent in this, which constitute MUST requirements (e.g. the implication of internationalization on the use of XML).
The basic assumptions are:
Security is here understood to be something else than privacy and trust. Security, in this context, is the safeguarding of the transmitted information in the course of transmission. These problems are usually addressed by the protocol used to carry the information, not the information structure or vocabulary itself. Privacy may be handled using the P3P mechanism of the W3C [P3P]. Integrity of the information is another matter, and is addressed in the requirements under section 4.1.3 below.
These requirements have been derived from the usecases above. But since several efforts exist to produce systems for content negotiation, we have also derived requirements from the work of others, especially the IETF Conneg working group (CONNEG) and the WAP Forum UAPROF Drafting Committee (UAP). Requirements have also been derived from the HTML working group in the W3C (XHTML), and from our own face-to-face meetings (F2F). Requirements have also been derived from experimental implementations presented to the working group (CCPP-IMPL).
The requirements are summarized in the order they impact the deliverables of the working group. However, requirements will typically impact the entire architecture, not one single deliverable. This listing is therefore for convenience only.
In addition, we have derived a set of protocol requirements. While the working group is not chartered to develop a protocol, it is still clear that its work will impact the design of any protocol used in the content negotiation process. These requirements have therefore been highlighted in Appendix 3.
The fulfillment of these requirements are mandatory for the design of CC/PP. They are MUST requirements.
Design goals are not as strong as requirements (SHOULD instead of MUST), and concern the entire architecture rather than a specific part of it.
The goals for the system are as follows:
Desirable features are not mandated, and optional to the design only if we feel like it. In terms of RFC 2119 [RFC2119] they are MAY-requirements.
FW-4: The framework shall allow for the description of preferences for its use, either through an algorithm within the framework or through a set of constraint semantics.
To fulfill the requirements in 3, 4 and 5, the CC/PP working group intend to produce a framework to handle and structure the profile information. The working group will also produce a basic vocabulary and describe how this can be extended (or other vocabularies can be created and referenced). In addition, the trust model for the work will be described in a separate document, as well as implementation issues.
The capabilities of the users client agent, and the preferences for the way the user wants them to be used, are expressed in a profile. The profile is constructed of a set of components that can be used to convey a group of attributes that are in some way related. The profile can consist of URI's that reference information or properties available in other documents or profiles.
The components that are available in a profile, along with the applicable attributes are specified as an instance of an RDF schema [RDF-Schema]. The profile may contain URI's that reference separately provided information or properties. A user agent or intermediary network entity can change the value of a CC/PP attribute. RDF is explained in further detail in section 6.2.
The foundation of RDF is a model for representing named properties and property values. The RDF model draws on principles from various data representation communities. RDF properties may be thought of as attributes of resources and in this sense correspond to traditional attribute-value pairs. RDF properties also represent relationships between resources and an RDF model can therefore resemble an entity-relationship diagram. (More precisely, RDF Schemas which are themselves instances of RDF data models are ER diagrams.) In object-oriented design terminology, resources correspond to objects and properties correspond to instance variables.
The RDF data model is a syntax-neutral way of representing RDF expressions. The data model representation is used to evaluate equivalence in meaning. Two RDF expressions are equivalent if and only if their data model representations are the same. This definition of equivalence permits some syntactic variation in expression without altering the meaning. (See Section 6.6 for additional discussion of string comparison issues.)
The basic data model consists of three object types:
RDF properties may be thought of as attributes of resources and in this sense correspond to traditional attribute-value pairs. RDF properties also represent relationships between resources. As such, the RDF data model can therefore resemble an entity-relationship diagram. The RDF data model, however, provides no mechanisms for declaring these properties, nor does it provide any mechanisms for defining the relationships between these properties and other resources. That is the role of RDF Schema.
Each RDF schema is identified by its own static URI. The schema's URI can be used to construct unique URI references for the resources defined in a schema. This is achieved by combining the local identifier for a resource with the URI associated with that schema namespace. The XML representation of RDF uses the XML namespace mechanism for associating elements and attributes with URI references for each vocabulary item used.
Please refer to the Namespaces in XML [XML-name] document for a complete description of how namespaces can be constructed in XML/RDF.
An RDF schema can declare constraints associated with classes and properties. In particular, the concepts of domain and range are used in RDF schemas to make statements about the contexts in which certain properties "make sense". Although the RDF data model does not allow for explicit properties (such as an rdf:type property) to be ascribed to Literals (atomic values), we nevertheless consider these entities to be members of classes (e.g. the string "John Smith" is considered to be a member of the class rdfs:Literal.) We expect future work in RDF and XML data-typing to provide clarifications in this area.
An RDF model that violates any of the consistency constraints specified in this document is said to be an inconsistent model. The following constraints are specified: rdfs:domain and rdfs:range constraints on property usage, the rule that rdfs:subPropertyOf and rdfs:subClassOf properties should not form loops, plus any further consistency constraints defined using the rdfs:ConstraintResource extensibility mechanism. Different applications may exhibit different behaviors in the face of an inconsistent model.
Some examples of constraints include:
RDF schemas can express constraints that relate vocabulary items from multiple independently developed schemas. Since URI references are used to identify classes and properties, it is possible to create new properties whose domain or range is constrained to be a class defined in another namespace.
The RDF Schema uses the constraint properties to constrain how its own properties can be used. These constraints are shown below in figure 7. Nodes with bold outlines are instances of rdfs:Class.
Please refer to the RDF Schema [RDF-schema] for a more complete description of RDF Constraints.
A CC/PP profile describes client capabilities in terms of a number of "CC/PP attributes", or "features". Each of these features is identified by a name in the form of a URI. A collection of such names used to describe a client is called a "vocabulary".
CC/PP defines a small, core set of features that are applicable to wide range of user agents, and which provide a broad indication of a clients capabilities. This is called the "core vocabulary". It is expected that any CC/PP processor will recognize all names in the core vocabulary, together with an arbitrary number of additional names drawn from one or more "extension vocabularies".
When using names from the core vocabulary or an extension vocabulary, it is important that all system components (clients, servers, proxies, etc.) that generate or interpret the names all apply a common meaning to the same name. It is preferable that different components use the same name to refer to the same feature, even when they are part of different applications, as this improves the chances of effective interworking across applications that use capability information.
Within an RDF expression describing a device, a vocabulary name appears as the label on a graph edge linking a resource to a value for the named attribute. The attribute value may be a simple string value, or another resource, with its own attributes representing the component parts of a composite value.
+-------------+ +------------+ | Resource |---attribute-name--->| Attribute | | | | value | +-------------+ +------------+
Simple string values may be used in comparison constraints [Ref-CCPP-format-document], which may interpret the attribute value as a textual or numeric value.
Vocabulary extensions are used to identify more detailed information than can be described using the core vocabulary. Any application or operational environment that uses CC/PP may define its own vocabulary extensions, but wider interoperability is enhanced if vocabulary extensions are defined that can be used more generally; e.g. a standard extension vocabulary for imaging devices, or voice messaging devices, or wireless access devices, etc.
Any CC/PP expression can use terms drawn from an arbitrary number of different vocabularies, so there is no restriction caused by re-using terms from an existing vocabulary rather then defining new names to identify the same information.
As indicated above, CC/PP attribute names are in the form of a URI. Any CC/PP vocabulary is associated with an XML namespace, which combines a base URI with a local XML element name (or XML attribute name) to yield a URI corresponding to an element name. Thus, CC/PP vocabulary terms are constructed from an XML namespace base URI and a local attribute name; e.g. the namespace URI:
and the core vocabulary name:
are combined to yield the attribute name URI:
Anyone can define and publish a CC/PP vocabulary extension (assuming administrative control or allocation of a URI for an XML namespace). For such a vocabulary to be useful, it must be interpreted in the same way by communicating entities. Thus, use of an existing extension vocabulary is encouraged wherever possible, or publication of a new vocabulary definition containing detailed descriptions of the various CC/PP attribute names.
Many extension vocabularies will be drawn from existing applications and protocols; e.g. WAP UAPROF, IETF media feature registrations, etc.
CC/PP expresses the user agent capabilities and how the user wants to use them.
XHTML document profiles[XHTML-docprof] express the required functionalities for what the author perceives as optimal rendering, and how the author wants them to be used.
The same mechanisms can be used to describe the document profile as the device capabilities profile, i.e. the framework described in section 6.2 combined with an extension vocabulary as described in 6.3.1 expressing the specific functionalities. This can, for instance, reflect the XHTML modules used, copyright issues, etc.
It is quite conceivable that when a device profile is expressed in CC/PP, it will have to be matched with a document profile expressed in a different format to achieve the adaption of content described above. This means that either of the profiles has to be translated to either the format of the other, or both to a common format.
In the scope of this activity, we will not discuss the translation of CC/PP profiles to other formats. We will regard the CC/PP format as the common format, to which other profile formats have been mapped (other groups are of course welcome to create converse mappings; we are trying to work out how this would work in the scope of our activity).
The interactions would work as follows:
This group will also describe how to map a different format to the CC/PP format, using the Conneg vocabulary as an example.
[CC/PP] Composite Capability/Preference Profiles (CC/PP): A user side framework for content negotiation
[CCPPex] CC/PP exchange protocol based on HTTP Extension Framework
[FAX-conneg] Content Negotiation for Facsimile Using Internet Mail
[HTTPex] HTTP Extension Framework
[HTTP/1.1] HTTP/1.1, Rev6
[Dsig] XML-Signature Syntax and Processing
[P3P] Platform for Privacy Preferences P3P Project
[P3P-Syntax] Platform for Privacy Preferences (P3P1.0) Syntax Specification
[RDF] Resource Description Framework, (RDF) Model and Syntax Specification
[RDF-Schema] Resource Description Framework (RDF) Schema Specification
[RFC2543] RFC 2543 : Media Features for Display, Print, and Fax
[RFC2506] RFC 2506 : Media Feature Tag Registration Procedure
[RFC2533] RFC 2533 : A Syntax for Describing Media Feature Sets
[RFC2703] RFC 2703 : Protocol-independent content negotiation framework
[RFC2738] RFC 2738 : Corrections to 'A syntax for describing media feature sets'
[RFC2246] RFC2246 : The TLS Protocol Version 1.0
[RFC2119] RFC 2119 : Key words for use in RFCs to Indicate Requirement Levels
[RFC2045] RFC 2045 : Multipurpose Internet Mail Extensions(MIME) Part One: Format of Internet Message Bodies
[RFC2396] RFC 2396 : Uniform Resource Identifiers (URI): Generic Syntax
[UAProf] User Agent Profiling Specification 10-Nov-1999
[WSP] WAP Wireless Session Protocol Specification
[WTP] WAP Wireless Transaction Protocol Specification
[WML] WAP Wireless Markup Language Specification
[WTLS] WAP Wireless Transport Layer Security Specification
[WBXML] WAP Binary XML Content Format Specification
[XHTML-docprof] XHTML Document Profile Requirements - Document profiles - a basis for interoperability guarantees
[XML] Extensible Markup Language (XML) 1.0
[XML-name] Namespaces in XML
[XSLT] XSL Transformations
This document has been edited by the editors, but the real credit goes to the CC/PP working group, especially those members who provided input to this document, listed below:
Anne Owen, Nortel
Barry Briggs, Interleaf
Chris Woodrow, Information Architects
Franklin Reynolds, Nokia
Graham Klyne, Content Technologies Ltd.
Hidetaka Ohto, W3C/Panasonic
Johan Hjelm, W3C/Ericsson. Working Group Chair.
Kynn Bartlett, HTML Writers Guild
Lalitha Suryanarayana, SBC Technology Resources
Mikael Nilsson, Ericsson. Editor.
Noboru IWAYAMA, Fujitsu
Sandeep Singhal, IBM. Chair WAP Forum UAProf DC.
Ted Hardie, Equinix
Ted Wugofski, Gateway2000
Ulrich Kauschke, T-Mobil
This is a listing of design assumptions, design goals, and requirements used in our work to derive the requirements in section 4 of the document.
(Note: The CONNEG goals are in an unordered list in the original document)
This section describes the requirements the CC/PP working group have been able to identify that applies to the protocol used to transport the information and/or perform the negotiation between the client and origin server. Since the CC/PP framework itself is intended to be independent from protocols, the protocol which conveys CC/PP information could be based on or extend a protocol of any kind such as HTTP, SMTP, LDAP etc.
The CC/PP information can be transported in in other messaging formats, for instance included in an email, sent on a backchannel, or out of band (either the head or the body) for describing the required capabilities of the recipient. How to convey content negotiation information in the extended MIME headers is being worked on the Internet FAX Working Group in the IETF [FAX-conneg].
The use case that is most likely (given that 80% of all Internet traffic is HTTP), is the use case where a web user agent sends a request with CC/PP information, and an origin server or intermediaries creates or selects tailored content, and includes the tailored content in the response. The use of HTTP has been a design assumption in the work.
Given this, the protocol requirements are as follows.
Assuming a protocol based on HTTP 1.1, the following applies: