Copyright © 2002 W3C ® (MIT, INRIA, Keio), All Rights Reserved. W3C liability, trademark, document use, and software licensing rules apply.
The World Wide Web is a networked information system. Web Architecture consists of the requirements, constraints, principles, and design choices that influence the design of the system and the behavior of agents within the system. When followed, the large-scale effect is that of a shared information space. This document organizes the technical discussion of the system in three parts: identification, representation, and interaction. This document also addresses some non-technical (social) issues that contribute to the shared information space.
This document strives to establish a reference set of requirements, constraints, principles, and design choices for Web architecture.
This section describes the status of this document at the time of its publication. Other documents may supersede this document. The latest status of this document series is maintained at the W3C.
This is the 15 November 2002 draft of "Architecture of the World Wide Web." This document has been developed by W3C's Technical Architecture Group (TAG) (charter). A list of changes in this document is available.
This draft remains incomplete; sections 1 and 2 are the most developed, 3 and 4 the least. The TAG has published a number of findings that address specific architecture issues. Parts of those findings may appear in subsequent drafts. Please also consult the list of issues under consideration by the TAG.
This draft includes some editorial notes and also references to open TAG issues. These do not represent all open issues in the document. They are expected to disappear from future drafts.
Publication as a Working Draft does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than "work in progress."
The latest information regarding patent disclosures related to this document is available on the Web. As of this publication, there are no disclosures.
Please send comments on this document to the public W3C TAG mailing list www-tag@w3.org (archive).
A list of current W3C Recommendations and other technical documents can be found at the W3C Web site.
The World Wide Web (or, Web) is a networked information system consisting of agents (programs acting on behalf of another person, entity, or process) that exchange information.
This document organizes Web architecture into:
The terms MUST, MUST NOT, SHOULD, SHOULD NOT, and MAY are used in accordance with RFC 2119 [RFC2119].
The intended audience for this document includes:
The authors have made every effort to keep this document terse, with the expectation that additional documents will elaborate on the required properties, constraints, and principles, rationale, and examples.
Readers will benefit from familiarity with the Requests for Comments (RFC) series from the IETF, some of which define pieces of the architecture discussed in this document.
This document focuses on the architecture of the Web. For instance, the principles enumerated in this document are those closely related to the Web. General design principles are not discussed in detail, such as minimal constraint (fewer rules makes the system more flexible), modularity, minimum redundancy, extensibility, simplicity, and robustness.
Other groups within W3C are addressing architectural design goals in the following areas:
For information about architectural principles of the Internet, refer to [RFC1958].
In the design of the Web, some design decisions, like the names
of the <p> and <li> elements in HTML, or the choice of
the colon character in URIs, are somewhat arbitrary; if
<par>, <elt>, or *
had been chosen
instead, the large-scale result would, most likely, have been the
same. Other design choices are more fundamental; these are the
focus of this document.
The terms used in the following list are elaborated on in the document.
Some of the items in the above list may conflict with current practice, and so education and outreach will be required to improve on that practice. Other items may fill in gaps in published specifications or may call attention to known weaknesses in those specifications.
The architecture described in this document is the result of experience. There has been some theoretical and modeling work in the area of Web Architecture, notably Roy Fielding's work on "Representational State Transfer" [REST].
The Web is a universe of resources. A resource is defined by [RFC2396] to be anything that has identity. Examples include documents, files, menu items, machines, and services, as well as people, organizations, and concepts. Web architecture starts with a uniform syntax for resource identifiers, so that we can refer to resources, access them, describe them, and share them. The Uniform Resource Identifier (URI) syntax employs an extensible set of URI schemes. Several URI schemes incorporate identification mechanisms that pre-date the Web into this (generic URI) syntax:
mailto:nobody@example.org
ftp://example.org/aDirectory/aFile
news:comp.infosystems.www
tel:+1-816-555-1212
Other URI schemes have been introduced since the advent of the Web, including those introduced as a consequence of new protocols. Examples of URIs for these schemes include:
http://www.example.org/something?with=arg1;and=arg2
ldap://ldap.itd.umich.edu/c=GB?objectClass?one
urn:oasis:SAML:1.0
One can append a fragment identifier to a URI to yield an identifier for part of, or a view of, a resource2. The following URIs include fragment identifiers:
ftp://example.org/aDirectory/aDocument#section1
http://www.example.org/states#texas
Note that while this composition is syntactically fully general,
it is meaningless in some URI schemes. The URI
mailto:nobody@example.org#abc
is meaningless in
practice.
A generic syntax for URIs is defined by [RFC2396]. The current document uses the term "URI" to mean, in RFC2396 terms, an absolute URI reference3 optionally followed by a fragment identifier. The TAG is working actively to convince the IETF to revise RFC2396 so that the definition of "URI" aligns with the current document.
When one resource refers to another via a URI, a link is formed. When many resources are linked this way, the large-scale effect is a shared information space, where resources are identifiable by URI. The value of the Web increases with the number of resources identified by URI; this is due to the "network effect." In turn, resources are more valuable when they are identifiable on the Web. Hence:
Constraint
Use URIs: All important resources SHOULD be identified by a URI.4
There are many benefits to making resources identifiable by URI. Some are by design (e.g., linking and bookmarking), while others have arisen naturally (e.g., global search services). See the TAG finding URIs, Addressability, and the use of HTTP GET for some details about the interaction of this principle in HTTP application design.
The two primary uses of URIs are:
There may be applications (e.g., XML namespace names [XMLNS]) where comparison is expected to be the sole or primary use of a URI. Certain URI schemes provide rules for determining the syntactic equivalence of URIs, i.e., whether two URIs are different spellings of the same identifier. These rules vary from scheme to scheme.
For example, URNs begin with two colon-delimited fields, the
first of which is the string urn
and the second is the
"namespace identifier" (NID). In URNs, these two
fields are to be compared in a case-insensitive fashion. The
remainder of the URN following the second colon is subject to rules
dependent on the content of the second field (following the first
colon) - thus the equivalence rules may vary within URN namespace
identifiers.
Section 3.2.3 of the HTTP specification [RFC2616] states that, when comparing two HTTP
URIs, the host name part must be considered case-insensitive, so
http://WWW.EXAMPLE/
and
http://www.example/
identify the same resource.
Good practice
URI case: It SHOULD NOT be assumed that URIs which differ only in character case can be used interchangeably.
Note: Equivalence of URIs is not the same as consistent representations of a resource.
Issue: URIEquivalence-15: When are two URI variants considered equivalent? See also issue IRIEverywhere-27 - Should W3C specifications start promoting IRIs?
To dereference a URI is to apply in succession a finite set of relevant specifications, beginning with the specification that governs the scheme of the URI.
A "representation" is a data object that represents or describes a resource state, and is the vehicle for conveying the meaning of a resource. A resource is an abstraction for which there is a conceptual mapping to a (possibly empty) set of representations.
As an example of the application of specifications in
succession, suppose that
http://weather.yahoo.com/forecast/MXOA0069
is used within an
a
element of an SVG document. The sequence of
specifications applied is:
a
link involves retrieving a representation of a
resource, identified by the XLink href
attribute: "By
activating these links (by clicking with the mouse, through
keyboard input, and voice commands), users may visit these
resources." This means that the GET method defined in HTTP/1.1 is
used to retrieve the representation of the resource.Representations, when transferred by a Web protocol, are often accompanied by metadata in the message (for example, HTTP headers). In particular, the value of the media type in the set of metadata is key to the correct interpretation of a resource representation, and governs the handling of fragment identifiers. See section 2 for more information about formats used to encode representations.
Depending on the protocol used, there may be several ways to dereference a URI. One of the most important actions on the Web is to retrieve a representation of a resource (such as with HTTP GET), which means to retrieve a representation of the state of the resource. There are other ways to interact with a resource (such as with HTTP POST). Dereference mechanisms vary by URI scheme. For instance, the URN scheme [RFC 2141] does not specify a dereference procedure.
Good practice
Resource descriptions: Owners of important resources SHOULD make available representations that describe the nature and purpose of those resources.
Issue: namespaceDocument-8: What should a "namespace document" look like?
Principle
Safe retrieval: Agents do not incur obligations by retrieving a representation.
For instance, a user does not incur an obligation by following an HTML link that causes the user agent to retrieve a representation. Tools such as proxies and search engines can retrieve representations without user interaction; it would be harmful to the Web if such operations incurred obligations. See the TAG finding "URIs, Addressability, and the use of HTTP GET" for more information about safe retrieval.
Issue: deepLinking-25: What to say in defense of principle that deep linking is not an illegal act?
URIs represent a worldwide contract for who can create names and how the resources they designate take on meaning. In the case of HTTP URIs, for example, the agreement is that the authoritative meaning of the resource designated by the URI is established by retrieving a representation of the resource (per the HTTP specification [RFC2616]) and then interpreting the representation according to the relevant specifications. The authoritative meaning of a resource is established by following specifications.
Representations of a resource may vary as a function of factors
including time, the identity of the agent accessing the resource,
data submitted to the resource when interacting with it, and
changes external to the resource. Consider the previous URI
http://weather.yahoo.com/forecast/MXOA0069
: representations
for the designed resource (the weather in Oaxaca) depend on (at
least) time, the expressed preference of the user for Fahrenheit or
Celsius, the identity of the user-agent software receiving the
representation, and, presumably, the weather in Oaxaca.
Good practice
Consistent representations: It is confusing and costly when, for a given URI, representations vary in unpredictable ways.
For example, serving two images as equivalents through HTTP content negotiation, where one image represents a square and the other a circle, will undermine confidence in the URI used to retrieve those images.
A description of what a URI identifies should be unambiguous.
For instance, saying that the URI
http://www.example.com/moby
identifies "Moby Dick" can lead
to confusion because this might be interpreted as any one of the
following very distinct resources: a particular printing of this
work (say, by ISBN), or the work itself in an abstract sense (for
example, using RDF), or the fictional white whale, or a particular
copy of the book on the shelves of a library (via the Web interface
of the library's online catalog), or the record in the library's
electronic catalog which contains the metadata about the work, or
the Gutenberg project's online
version. Similarly, one should not use the same URI to refer to
a person and to that person's mailbox. See issue
httpRange-14.
There are thus strong social expectations that once a URI identifies a particular resource, it should continue indefinitely to refer to that resource; this is called the persistence of the URI. Persistence is always a matter of policy and commitment on the part of authorities assigning URIs rather than a constraint imposed by technological means.
For example, each W3C technical report (e.g., "the SVG
specification") is in fact a series of documents that mature over
time (from Working Drafts, Candidate Recommendations, Proposed
Recommendations, to Recommendation). W3C assigns a URI to the
"latest version" in the series (e.g.,
http://www.w3.org/TR/SVG
). W3C also assigns a URI for each
specification in the series (called the "this version URI", e.g.,
http://www.w3.org/TR/2001/PR-SVG-20010719/
). W3C
policy is that representations of the "latest version" resource
will change over time (with each new publication of an SVG
specification). W3C policy is also that representations of a
specification designated by a "this version" identifier will not
change over time, to the best of W3C's ability to maintain its
archives intact.
HTTP [RFC2616] has been designed to promote consistency. For example, HTTP redirection (via some of the 3xx response codes) permits servers to tell a client that further action needs to be taken by the client in order to fulfill the request (e.g., the resource has been assigned a new URI). In addition, content negotiation also promotes consistency, as a site manager would not be required to define new URIs for each new format that is supported, as would be the case with protocols that don't support content negotiation, such as FTP.
For more discussion about persistence, refer to [Cool].5
It is confusing and costly when people use the same URI to refer
to different resources (i.e., where there is some inconsistency in
usage compared to the authoritative meaning of the resource).
Suppose company A uses http://example.com/coolcompany
to refer to CoolCompany's home page, while company B uses
http://example.com/coolcompany
to refer to CoolCompany.
Company A then buys company B, but when they try to merge their
databases, they cannot due to this inconsistent usage of the
URI.
Good practice
Consistent URIs: Indiscriminate use of a URI undermines its value and interferes with people who rely on it.
One
important characteristic of a URI is its
scheme (the string that precedes the first colon in a
URI). For example the scheme of the URI
http://www.example.com/
is "http", and for
ftp://ftp.example.com/
it is "ftp". It is common to classify
URIs by scheme, calling the two preceding examples respectively an
"HTTP URI" and an "FTP URI".
Since many aspects of URI processing are scheme-dependent, and since a huge range of software is expected to be able to process URIs, the cost of introduction of new URI schemes is very high.
Good practice
New URI schemes: Authors of specifications SHOULD avoid introducing new URI schemes when existing schemes can be used to meet the goals of the specifications.
While "myscheme:blort" is a URI that satisfies the syntactic constraints of [RFC2396], if "myscheme" is not registered, you are not guaranteed that somebody else isn't already using it for something else.
The IANA registry [IANASchemes] lists registered URI schemes and the specifications that define them. For instance, the IANA registry indicates that the "http" scheme is defined by [RFC2616]. Refer to RFC2717 for information about registering a new URI scheme.
The deployment and use of different URI schemes may require varying degrees of central coordination and administration. For example, MAILTO, FTP, and HTTP URIs depend (in practice at least) on the use of the DNS infrastructure. Also, there is a central registry of URN namespace identifiers.
In some URI schemes it is meaningful for a URI to end with a fragment identifier. The fragment identifier is interpreted only after the retrieval of a representation. Section 4.1 of [RFC2396] states that "the format and interpretation of fragment identifiers is dependent on the media type [RFC2046] of the retrieval result," that is, the representation.
For instance, if the representation is an HTML document, the fragment identifies a hypertext anchor. In the case of a graphics format, the fragment might identify a circle or spline. In the Resource Description Framework [RDF10], fragments can be used to identify anything, be it abstract (e.g., a dream) or concrete (e.g., an automobile).
Good practice
Coneg with fragments: Authors SHOULD NOT use HTTP content negotiation for different media types that do not share the same fragment identifier semantics.
Editor's note: There has been some discussion but no agreement that new access protocols should provide a means to convert fragment identifiers according to media type.
The following generalities about URIs are included to answer some frequently asked questions about URIs. These are generalities because they hold for some, but not necessarily all, URI schemes.
http://www.example.com/lj45sr
and know that it refers to "my
old car" or "the weather forecast for Oaxaca."
Over time, we trust that some URIs will identify familiar resources, but that trust derives from social behavior, not the spelling of the identifier.
Data on the Web manifests itself through resource representations. A resource representation consists of:
A format specification describes the structure of the bit sequence.
Refer to other W3C format guidelines: Charmod, XAG, etc.
What is a format, and how does it relate to the concept of a document. Do all documents have a format? Is a document a collection of resources of different formats organized into a whole? Is a document the same as a resource? the same as a message body? as a non-multipart message body? What is the distinction between documents and data, if any. Does 'document' imply human readable and if so, does it imply presentation? Does it imply a hierarchically structured, report-like document with headings and subheadings? Is a catalog a document? Is a rave flyer a document?
Negotiation (stuff above might go here also) by network request, by listed alternatives in content any preference? Resource variants, foo.css and foo.html unlikely to be equivalent.
On the interpretation and processing of formats (see namespaceDocument-8 and mixedNamespaceMeaning-13):
@@Incomplete sections on specification design.@@
On using XML:
This section attempts to organize some areas of future discussion. Separating the concepts content, presentation, and interaction allows more easily composable specifications. For example, a markup language can be specified independently of a style sheet language. The separation facilitates alternate presentations of the same content, which is seen to have an accessibility advantage and to be more suited to the multiple modalities of Web access.
Issue: contentPresentation-26: Separation of semantic and presentational markup, to the extent possible, is architecturally sound.
Composability (ns-meaning). Use of XML for tree structured content. Linking in general v. idref in one document. Human readable v. machine data. Served or not (hidden behind server - semantic firewall, accessibility. Linking into parts of the content, transclusion of parts. Compound documents, components from multiple servers - scalability, deep linking. Processing models, error handling.
Presentation by decoration (application of CSS to XML as presentation), and by derivation (creation of html/svg/etc as presentation). Linking (bidirectionally) between content and presentations. Inheritance of properties across namespaces. Consistency of property names. Subsets. 'Applies to' as opposed to 'set on'. Specificity of properties as attributes, chaining styling, restyling. Time-lines, linking to portions of a time-line.
Animation, scripting, events, client/server interaction. Declarative v. script based - accessibility, power; formalization of common functionality (loop animation, rollovers) in declarative form. DOM - making additional methods, add to rather than replacing XML DOM. Effect of script/programming language limitations on choice of element and attribute names. Linking to active components - XForms example with model and abstract form control, can be extended to presentational instantiation of form control.
As mentioned in the introduction, the Web is designed to create the large-scale effect of a shared information space that scales well and behaves predictably.
@@There may be some general principles that hold across all three previous chapters. Put them in an appendix and refer to them from each section?@@
When designing specifications that address independent functions of a system, avoidable references between the specifications are in general harmful. They are harmful because they impede the independent evolution of the specifications.
For example, it is a strength of XML that XPath cannot query the
HTTP header. It is a strength of HTTP that it does not refer to
details of the underlying TCP to the extent that it cannot be run
over a different transport service. Similarly, the RDF data graph
has a significance that is independent of the actual serialization.
However, there is a flaw: the embedded XML
parsetype="Literal"
data type.
Sometimes it is necessary (and good for given application) to break layers. For example, it is good for an HTTP client to be aware of TCP speeds and round trip times to different mirror servers in order to optimize the choice of server. When designing specification, identify the functionalities that break layers so it is clear when they are being used.
http://example/dir1/dir2/file1
, the relative URI reference
../file2
is a shortened form of
http://example/dir1/file2
and the relative URI reference
#abc
is a shortened form for
http://example/dir1/dir2/file1#abc
. (Note 3
context.)The authors of this document are the participants of W3C's Technical Architecture Group: Tim Berners-Lee (Chair, W3C), Tim Bray (Antarctica Systems), Dan Connolly (W3C), Paul Cotton (Microsoft), Roy Fielding (Day Software), Chris Lilley (W3C), David Orchard (BEA Systems), Norman Walsh (Sun), and Stuart Williams (Hewlett-Packard).
The TAG thanks people for their thoughtful contributions on the TAG's public mailing list, www-tag (archive).