XML Linking

October 2, 1997

Steve J. DeRose and Tim Bray

Extensible Markup Language (XML) Part 2: Linking

Tim Bray, Steve DeRose

Abstract

[W3C Working Draft; WD-xml-link-970731; July 31, 1997]

This document specifies a simple set of constructs that may be inserted into XML documents to describe links between objects and to support addressing into the internal structures of XML documents. It is a goal to use the power of XML to create a structure that can describe the simple unidirectional hyperlinks of today's HTML as well as more sophisticated multi-ended, typed, self-describing links.

Status of This Memo

This is a W3C Working Draft for review by W3C members and other interested parties. It is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to use W3C Working Drafts as reference material or to cite them as other than "work in progress." A list of current W3C working drafts can be found at http://www.w3.org/TR.

Note that some of the work that is still be done in this particular draft is described in Appendix A.This work is part of the W3C SGML Activity (for current status, see http://www.w3.org/MarkUp/SGML/Activity).

1. Introduction

This document specifies a set of constructs which may be inserted in XML documents to describe links. A link, as the term is used here, is a relationship which is asserted to exist between two or more data objects or portions of data objects. This specification is concerned with the syntax used to assert link existence and describe link characteristics. Implicit (unasserted) relationships, for example that of one word to the next, or that of a word in a text to its entry in an on-line dictionary, are outside its scope. Explicitly asserted links do not constitute the only useful kind of link, but this specification is intended neither to provide machinery for every possible kind of link nor to preclude the use of such machinery.

The existence of links is asserted by the presence of elements contained in XML documents. They may or may not reside at the locations of, or in the same documents with, the objects which they serve to connect.

1.1 Origin and Goals

This specification aims to provide an effective, compact structure for representing links that can be in documents or external to them, and that can have multiple typed locators, indirection, and precise specification of resource locations in XML and SGML data. It also aims to represent the abstract structure and significance of links, leaving formatting issues to stylesheets or other mechanisms as far as is practical.

1.2 Relationship to Existing Standards

Three standards have been especially influential:

HTML

Defines several SGML element types that represent links as well as popularizing a location specifier type, the URL, mainly focused on pointing to entire data objects, though with some provision for linking to elements with IDs, regions in graphics, and so on.

HyTime

Defines location specifier types applicable to all kinds of data, as well as in-line and out-of-line link structures, and some semantic features, including traversal control and placement of objects into a display or other space.

Text Encoding Initiative Guidelines (TEI P3)

Provides a formal syntax for location specifiers for structured data, graphics, and other data, and structures for creating links and link collections out of them.

Many other linking systems have also informed this design, including Dexter, MicroCosm, and InterMedia.

1.3 Terminology

The following basic terms apply in this document:

resource

In the abstract sense, an addressable unit of information or service which is participating in a link. Examples include files, images, documents, programs, and query results. Concretely, anything which happens to be reachable by the use of a locator in some linking element. Note that this term and its definition are taken from the basic specifications governing the World Wide Web.

linking element

An element which asserts the existence and describes the characteristics of a link.

locator

A character string appearing in a linking element, which may be used to locate a resource.

title

A caption associated with a resource, suitable for showing users as a means of explaining the significance of the part played in the link by that resource.

traversal

The action of using a link; that is, of accessing a resource. Traversal may be initiated by a user action (for example, clicking on the displayed content of a linking element) or occur under program control.

multi-directional link

A link that can be traversed starting at more than one of its resources. Note that being able to "go back" after following a one-directional link does not make the link multi-directional.

Abstractly, a link which serves as one of its own resources. Concretely, in the language of this specification, a link where the content of the linking element serves as a resource. HTML A, HyTime clink, and TEI XREF are all examples of in-line links.

A link which does not serve as one of its own resources. Such links only make sense given a notion like link groups, which instruct applications where to look for links. Nevertheless, out-of-line links are required to support multi-directional traversal and for creating links with resources which can be traversed starting from read-only data objects.

1.4 Notation

The formal grammar for locators is given using a simple Extended Backus-Naur Form (EBNF) location, as described in Part 1.

1.5 Types of Link Types

There is an extensive literature on link typology. Some well-known axes are:

link relationships

Links express various kinds of relationships between the data objects or portions they connect, in terms of conceptual significance to the author and user. Some links may be criticisms, others add support or background, while others have a very different meaning such as providing access to demographic information about a data object (its author's name, version number, etc.), or to navigational tools such as index, glossary, and summary.

In-line and out-of-line links differ in their structure, as do links involving varying numbers of resources.

There are many languages available for use in locating resources; examples are URLs, SQL queries, and and file names.

Links may be presented in a variety of ways. The discussion of this area is complicated by the fact that link formatting and link behavior are inextricably connected. This specification does not discuss, or provide mechanisms for, the provision or use of link formatting information.

Links may have a wide variety of effects when traversed, such as opening, closing, or scrolling windows or panes; displaying the data from various resources in various ways; testing, authenticating, or logging user and context information; or executing various programs. Ideally, link behavior should be determined by a semantic specification based on link types, resource roles, user circumstances, and other factors; just as element formatting is determined by a stylesheet based on element type, context, and other factors.

2. Link Recognition

2.1 Linking Element Recognition

The existence of a link is asserted by a linking element. Linking elements must be recognized reliably by software in order to provide appropriate display and behavior. XML linking elements are recognized based on the use of a designated attribute named XML-LINK. Possible values are SIMPLE, EXTENDED, LOCATOR, GROUP, and DOCUMENT, signalling in each case that the element in whose start-tag the attribute appears is to be treated as an element of the indicated type, as described in this specification.

An example of such a link follows:

 <A XML-LINK="SIMPLE" HREF="http://www.w3.org/"The W3C</A> 

2.2 Attribute Remapping

This specification describes many attributes that can be attached to linking elements to describe various aspects of links. Each is given a name in this specification. It may be desired to use existing elements in XML documents as linking elements, but such elements might already have attributes whose names conflict with those described in this document. To avoid collisions, user-chosen attributes can be declared as equivalent to those described in this specification using the XML-ATTRIBUTES attribute.

This attribute must contain an even number of white-space-separated names, which are treated as pairs. In each pair, the first name must be one of those described in this specification: (ROLE, HREF, TITLE, SHOW, INLINE, CONTENT-ROLE, CONTENT-TITLE, ACTUATE, BEHAVIOR, STEPS). The second name, when recognized in the document, will be treated as though it were playing the role assigned to the first. For example, consider a DTD with the declaration shown in Example 1.

EXAMPLE 1

 
        <!ELEMENT TEXT-BOOK ANY> 
        <!ATTLIST TEXT-BOOK TITLE CDATA #IMPLIED 
                               ROLE (PRIMARY|SUPPORTING) #IMPLIED> 

If it were desired to use this as a simple link, it would be necessary to remap a couple of attributes, which could be accomplished in the internal subset shown in Example 2.

EXAMPLE 2

 
        <!ATTLIST TEXT-BOOK XML-LINK CDATA #FIXED "SIMPLE" 
                               XML-ATTRIBUTES CDATA #FIXED 
                                                    "TITLE XL-TITLE ROLE XL-ROLE"> 

Then in the document, Example 3 would be recognized as a simple link.

EXAMPLE 3

 
        <TEXT-BOOK TITLE="Compilers: Principles, Techniques, and Tools"
                      ROLE="PRIMARY" XL-TITLE="Primary Textbook for the Course"
                      XL-ROLE="ONLINE-PURCHASE"
                      HREF="/cgi/auth-search?q="+Aho+Sethi+Ullman"/> 

2.3 Operational Issues Concerning
Link Recognition

There are two distinct mechanisms that may be used to associate the XML-LINK and XML-ATTRIBUTES attributes with a linking element. The simplest is to provide it explicitly. However, this practice is verbose, and would be not only cumbersome but wasteful of network bandwidth in the case where there are large numbers of linking elements. Fortunately, XML's facilities for declaring default attribute values can be used to address this problem. For example, the following would accomplish the declaration of the A element as an XML SIMPLE link:

        <!ATTLIST A XML-LINK CDATA #FIXED "SIMPLE">
      

Such a declaration may be placed in either the external or the internal subset of the Document Type Declaration. Placing it in both subsets would be the obvious thing to do for convenient network operation. So doing, at the time of creation of this specification, would cause the document to fail to be valid. Note that the successful completion of the current work on a technical corrigendum to ISO 8879 that is in the process of international ballot would resolve this problem and allow this practice in valid documents. However, for interoperability, the declaration should not be placed in both subsets.

3. Linking Elements

This specification defines two types of linking elements. First, a simple link, which is usually in-line and always one-directional, very like the HTML A element. Second, a much more general extended link which may be either in-line or out-of-line and may be used for multi-directional links, links into read-only data, and so on.

3.1 Information Associated with Links

This specification describes a variety of information that may be (and in some cases is required to be) associated with linking elements:

Every link may have a role, a string used to identify to the application program the meaning of the link. Furthermore, each resource participating in a link may be given its own role. The ROLE attribute is used to provide both link and resource roles.

Every locator must identify a resource in some fashion. This is done using the HREF attribute, as described in section 5, "Addressing."

Every locator may be associated with a title in the TITLE attribute. This specification does not require that applications make any particular use of the title.

The SHOW and ACTUATE attributes may be used by an author to communicate general policies concerning the traversal behavior of the link; this specification defines a small set of policies for this purpose. The BEHAVIOR attribute may be used to communicate detailed instructions for traversal behavior; this specification does not constrain the contents, format, or meaning of this attribute.

The INLINE attribute may be used to communicate whether the linking element is in-line or not.

3.2 Content of Linking Elements

This specification places no constraints on the contents of linking elements; for this reason, in the sample declarations given below, they are given declared content of ANY.

Any element may be recognized as a linking element based on use of the XML-LINK attribute; in a valid document, each such element must conform to the constraints expressed in its governing DTD.

3.3 Simple Links

Simple links are very much like HTML A or TEI XREF elements, but with more general reference capabilities. A simple link may contain only one locator; thus there is no necessity for a separate child element, and the locator attributes are attached directly to the linking elements.

Example 4 is a sample declaration for an XML simple link; note that the element type need not be SIMPLE, since the linking element will be recognized based on the value of the XML-LINK attribute.

EXAMPLE 4

 
        <!ELEMENT SIMPLE ANY> 
        <!ATTLIST SIMPLE XML-LINK CDATA #FIXED "SIMPLE"
                            ROLE CDATA #IMPLIED 
                            HREF CDATA #REQUIRED 
                            TITLE CDATA #IMPLIED 
                            INLINE (TRUE|FALSE) "TRUE" 
                            CONTENT-ROLE CDATA #IMPLIED 
                            CONTENT-TITLE CDATA #IMPLIED 
                            SHOW (EMBED|REPLACE|NEW) "REPLACE" 
                            ACTUATE (AUTO|USER) "USER" 
                            BEHAVIOR CDATA #IMPLIED > 

3.4 Extended Links

An extended link can involve any number of resources, and need not be co-located with any of them. An application may be expected to provide traversal among all of them (subject to semantic constraints outside the scope of this paper). The key issue with extended links is how to manage and find them, since they do not necessarily co-occur with any of their resources, and often are located in completely separate documents. This process is discussed in Section 7, "Extended Link Groups."

A extended link's locators are contained in child elements of the linking element, each with its own set of attributes. Once again, in the sample declaration in Example 5, the element types need not be EXTENDED and LOCATOR; recognition depends on the XML-LINK attribute.

EXAMPLE 5

        <!ELEMENT EXTENDED ANY> 
        <!ELEMENT LOCATOR ANY> 
        <!ATTLIST EXTENDED XML-LINK CDATA #FIXED "EXTENDED" 
                     ROLE CDATA #IMPLIED 
                     TITLE CDATA #IMPLIED 
                     INLINE (TRUE|FALSE) "TRUE" 
                     CONTENT-ROLE CDATA #IMPLIED 
                     CONTENT-TITLE CDATA #IMPLIED 
                     SHOW (EMBED|REPLACE|NEW) "REPLACE" 
                     ACTUATE (AUTO|USER) "USER" 
                     BEHAVIOR CDATA #IMPLIED>
        <!ATTLIST LOCATOR 
                     XML-LINK CDATA #FIXED "LOCATOR" 
                     ROLE CDATA #IMPLIED
                     HREF CDATA #REQUIRED 
                     TITLE CDATA #IMPLIED 
                     SHOW (EMBED|REPLACE|NEW) "REPLACE"
                     ACTUATE (AUTO|USER) "USER" 
                     BEHAVIOR CDATA #IMPLIED >

The declared content of ANY for the linking element is perhaps misleading; the idea is that locator elements should appear as children of the linking elements, along with any other content that is appropriate.

Note that many of the attributes may be provided for both the parent linking element and the child locator element. If any such attribute is provided in the linking element but not in a locator element, the value provided in the linking element is to be used in processing the locator element. In other words, the attributes provided in the linking element may serve as defaults for the (possibly many) locator elements.

3.5 In-Line and Out-of-Line Links

The INLINE attribute can take the values TRUE and FALSE. The value TRUE, which is the default, means that all of the content of the linking element is to be considered a resource of the link, except for any child locator elements (which are considered part of the linking element machinery).

When the link is in-line, the CONTENT-ROLE and CONTENT-TITLE attributes may be used to provide the title and role information for this "content" resource. If INLINE is FALSE, it is not an error to provide the CONTENT-TITLE or CONTENT-ROLE attributes, but they have no effect.

4. Link Behavior

This specification provides a mechanism for the authors of linking elements to signal their intentions as to the timing and effects of traversal. Such intentions can be expressed along two axes, labeled SHOW and ACTUATE. These are used to express policies rather than mechanisms; programs which are processing links in XML documents are free to devise their own mechanisms, best suited to the user environment and processing mode, to implement the requested policies.

In many cases, there will be a requirement for much finer control over the details of traversal behavior; existing hypertext software typically provides such control. Such fine control of link traversal is outside the scope of this specification; however, the BEHAVIOR attribute is provided as a standard place for authors to provide, and in which programs should look for, such detailed behavioral instructions.

4.1 The SHOW Axis

The SHOW attribute is used to express a policy as to the context in which a resource that is traversed to should be displayed or processed. It may take one of three values:

EMBED

Directs that upon traversal of the link, the designated resource should be embedded, for the purposes of display or processing, in the body of the resource, and at the location, where the traversal started.

Directs that upon traversal of the link, the designated resource should, for the purposes of display or processing, replace the resource where the traversal started.

Directs that upon traversal of the link, the designated resource should be displayed or processed in a new context, not affecting that of the resource where the traversal started.

4.2 The ACTUATE Axis

The ACTUATE attribute is used to express a policy as to when traversal of a link should occur. It may take one of two values:

Directs that the link should be traversed when encountered, and that the display or processing of the resource where the traversal started is not considered complete until this is done.

Directs that the link should not be traversed until there is an explicit external request for this to happen.

5. Addressing

The locator value for a resource is provided in the HREF attribute. HREF may have at one point stood for "Hypertext reference"; the name is adopted for compatibility with existing practice.

5.1 Locator Syntax in General

A locator always contains a URL, as described in IETF RFCs 1738 [2] and 1808 [3]. As these RFCs state, the URL may include a trailing query (marked by a ?), and be followed by a # and a fragment identifier, with the query interpreted by the host providing the indicated resource, and the interpretation of the fragment identifier dependent on the data type of the indicated resource. Thus, when a locator in an XML linking element identifies a resource that is not an XML document (for example, an HTML or PDF document), this specification does not constrain the syntax or semantics of the query nor of the fragment identifier.

5.2 Locator Syntax for XML Resources

When a locator identifies a resource that is an XML document, the locator value may contain either or both a URL and a fragment identifier, which in this case is a TEI-derived "Extended Pointer" (hereinafter XPointer). Special syntax may be used to request the use of particular processing models in accessing the locator's resource. This is designed to reflect the realities of network operation, where it may or may not be desirable to exercise fine control over the distribution of work between local and remote processors.

Locator

        [1]   Locator  ::=  URL 
        | Connector (XPointer | Name) 
        | URL Connector (XPointer | Name) 
        [2] Connector  ::=  '#' | '|' 
        [3]       URL  ::=  URLchar* 
      

In this discussion, the term designated resource refers to the resource participating in the link which the locator serves to locate. The following rules apply:

• The URL, if provided, locates a resource called the containing resource.
• If the URL is not provided, the containing resource is considered to be the document in which the linking element is contained.
• If the XPointer is provided, the designated resource is a "sub-resource" of the containing resource; otherwise the designated resource is the containing resource.
• If the Connector is followed by a Name, the Name is shorthand for the XPointer "ID(Name)"; i.e., the sub-resource is the element in the containing resource that has an XML ID attribute whose value matches the Name. This shorthand is to encourage use of the robust ID addressing mode.
• If the connector is "#", this signals an intent that the containing resource is to be fetched as a whole from the host that provides it, and that the XPointer processing to extract the sub-resource is to be performed on the client, that is to say on the same system where the linking element is recognized and processed.
• If the connector is "|", no intent is signaled as to what processing model is to be used to go about accessing the designated resources.

In the case where the URL contains a query (to be interpreted by the server) information providers and authors of server software are urged to use queries as shown in Query Syntax:

6. Extended Pointers

XML uses a locator syntax derived from that for TEI extended pointers; the XML version of a TEI extended pointer is referred to as an XPointer. XPointers operate in a straightforward way on the element tree which is defined by the elements of an XML document.

The basic form of an XPointer is a series of location terms, each of which specifies a location, either absolute or (more frequently) relative to the prior one. Each term has a keyword such as ID, CHILD, ANCESTOR, and so on, and can be qualified by parameters such as an instance number, element type, or attribute.. For example, the locator string

        CHILD(2,CHAP)(4,SEC)(3)
      

refers to the third child of the fourth SEC within the second CHAP within the referenced document.

The syntax for TEI Extended Pointers has been adjusted in order to allow them to be packaged naturally with URLs without requiring URL-escaping of space characters:

• Parameters in locator terms must be separated by commas to facilitate including XPointers within URLs, where spaces would otherwise need to be escaped.
• A locator may contain two XPointers separated by the string "..". These define the beginning and end of a span which constitutes the resource. This merely combines the capability of the TEI FROM and TO attributes into the locator syntax.

6.1 XPointer Structure

At the heart of the XPointer is the location term. Location terms are designed to work in sequences and are explained fully below.

A locator can contain either one or two XPointers; if there are two, they are separated by the string "..". For a locator with one XPointer, the designated resource is the element or location selected by the sequence of location terms it contains. With two XPointers, the designated resource is all of the text from the location, or start of the element, selected by the first, through to the location, or the end of the element, selected by the second.

Note that the implementation of traversal to a resource is not constrained by this specification. In particular, handling a resource designated by a span is probably highly application-dependent. In a display-oriented application, such traversal might simply be implemented by highlighting the designated characters. In particular, it should be noted that a span cannot safely be treated as a set of elements; most spans will include partial elements.

A location term is an atomic unit of addressing information; XPointers consist of combinations of location terms. Location terms are grouped into absolute terms, relative terms, and string-match terms. Absolute terms select one or more elements or locations in an XML document; if an XPointer contains only an absolute term, that term identifies its designated resource. If the absolute term is followed by any relative or string-match terms, the elements or locations that it designates are termed a location source and serve as a starting point for the operations of the location terms in Absolute, Relative, and String-match.

6.2 Absolute Location Terms

The keywords described in this section do not depend on the existence of a location source. They can be used to establish a location source or as standalone self-contained XPointers.

If an XPointer omits any leading absolute location terms (i.e., consists only of relative and string-match terms) it is assumed to have a leading ROOT() absolute location term.

The empty parentheses after ROOT, HERE, and DITTO are for consistency with other keywords and to avoid ambiguous interpretation of an extended pointer containing just the string "ROOT" or "HERE".

6.2.1 The ROOT keyword

If an XPointer is preceded by ROOT(), the location source is the root element of the containing resource. This is the default behavior. ROOT keyword has no effect on the interpretation of the locator; it exists in the interests of design clarity.

6.2.2 The HERE keyword

If the first or second XPointer is preceded by HERE(), the location source for the first location term of that series is the linking element containing the locator rather than the default root element. This allows extended pointers to select items such as "the paragraph immediately preceding the one within which this pointer occurs." It is an error to use HERE in a locator where a URL is also provided and identifies a resource different from the document which contains the linking elements.

XPointer

Absolute, Relative, and String-match

Absolute Location Terms

6.2.3 The DITTO keyword

If the second XPointer is preceded by DITTO(), the location source for its first location term is the location source specified by the entire first XPointer in order to facilitate relative specification of a span.

6.2.4 The ID keyword

If the first or second XPointer is preceded by ID(Name), the location source for the first location term is the element in the containing resource which has an attribute of type ID with a value matching the given Name.

For example, the location specification

        ID(a27)
      

chooses the necessarily unique element of the containing resource which has an attribute declared to be of type ID whose value is a27.

6.2.5 The HTML keyword

The location term HTML(NAMEVALUE) selects the first element whose type is A and which has a NAME attribute whose value is the same as the supplied NAMEVALUE; this is exactly the function performed by the "#"-fragment in the context of an HTML document.

6.3 Relative Location Terms

The keywords described in this section and shown in Relative Location Terms depend on the existence of a location source. They provide facilities for navigating forward, backward, up, and down through the element tree.

The keyword selects zero or more elements relative to the location source, which are referred to as candidate locations. Each keyword summarized here is described in detail in following sections.

CHILD

Selects child elements of the location source.

Selects elements appearing within the content of the location source.

Selects elements in whose content the location source is found.

Selects elements which appear before the location source.

Selects preceding sibling elements of the location source.

Selects elements which appear after the location source.

Selects following sibling elements of the location source.

Relative Location Terms

Relative Location Term Arguments

6.3.1 Relative location term arguments

All relative location terms operate using arguments; each keyword takes the same set of arguments. Multiple sets of arguments can be attached to a single keyword as a shorthand for repeated occurrences of that keyword.

Multiple argument lists are a shorthand in which the keyword is considered to have been repeated between each of the steps. That is to say, the following two XPointers are equivalent:

        CHILD(2,SECTION)(1,SUBSECTION)
        CHILD(2,SECTION)CHILD(1,SUBSECTION)
      

6.3.2 Selection by Instance number

When the value of Instance is the number N, it selects the Nth of the candidate locations. If the special value ALL is given, then all the candidate locations are selected. Negative numbers count from the last candidate location to the first; numbers out of range constitute an error.

6.3.3 Selection by element type

Candidates can be selected by element type as well as number.

The ElType gives an XML element type; only elements of that type will be selected from among the candidate locations. For example, the location term

 CHILD(3,DIV1)(4,DIV2)(29,P) 

selects the 29th paragraph of the fourth sub-division of the third major division of the location source.

The XPointer

 DESCENDANT(-1,EXAMPLE)> 

selects the last example in the document.

Selection by type is strongly recommended because it makes links more perspicuous and more robust. It is perspicuous because humans typically refer to things by type: as "the second section," "the third paragraph," etc. It is robust because it increases the chance of detecting breakage if (due to document editing) the target originally pointed at no longer exists.

The type may be specified by Name or by using one of the values ".", "*CDATA", or "*". If the type is specified as ".", candidate elements of any type are matched. If the type is specified as "*CDATA", the location term selects only untagged sub-portions of an element with mixed content (these are generally referred to as pseudo-elements). Finally, * selects among child elements and pseudo-elements.

Consider the following example:

 
        <SPEECH ID="a27"><SPEAKER>Polonius </SPEAKER>
        <DIRECTION>crossing downstage </DIRECTION>Fare you well, 
         my lord. <DIRECTION>To Ros. </DIRECTION>
        You go to seek Lord Hamlet? There he is.</SPEECH>
      

Selects the second "DIRECTION" element, "To Ros."

Selects the second child element, which is the first direction "crossing downstage".

Selects the second pseudo-element (the line-break between the "SPEAKER" and "DIRECTION" elements is the first), "Fare you well, my lord."

Selects the second element or pseudo-element among the children, the line-break between the "SPEAKER" and "DIRECTION" elements.

Instance

Element Type

6.3.4 Selection by attribute

Candidates can be selected based on attribute name and value.

The Attr and Val are used to provide attribute names and values to use in selecting among candidates.

If specified within quotation marks, the attribute-value parameter is case-sensitive; otherwise not.

As with generic identifiers, attribute names may be specified as * in location terms in the (unlikely) event that an attribute value constitutes a constraint regardless of what attribute name it is a value for.

For example, the location term

 CHILD(1,*,TARGET,*) 

selects the first child of the location source for which the attribute TARGET has a value.

The location specification

 CHILD(1,*,N,2)(1,*,N,1) 

chooses an element using the N attribute. Beginning at the location source, the first child (whatever element type it is) with an N attribute having the value 2 is chosen; then that element's first child element having the value 1 for the same attribute is chosen.

The location specification

 CHILD(1,FS,RESP,*IMPLIED) 

selects the first child of the location source which is an FS element for which the RESP attribute has been left unspecified.

Note that the HTML keyword is a synonym for a very specific instance of attribute-based addressing such that the following two XPointers are equivalent:

 HTML(Sec3.2) ROOT()DESCENDANT(1,*,A,"Sec3.2") 

6.3.5 The DESCENDANT keyword

The location specification

 ID(a23)DESCENDANT(2,TERM,LANG,DE) 

selects the second TERM element with a LANG attribute whose value is DE occurring within the element with an ID attribute whose value is A23. The search for matching elements occurs in the same order as the XML data stream (depth-first, left-to-right).

If an instance number is negative, the search is depth-first right-to-left, in which the right-most, deepest matching element is numbered -1, etc. The location specification

Attribute

        ROOT()DESCENDANT(-1,NOTE)
      

thus chooses the last NOTE element in the document, that is, the one with the rightmost start-tag.

6.3.6 The ANCESTOR keyword

The ANCESTOR location term selects an element from among the direct ancestors of the location source. The parameters are for CHILD. However, the ANCESTOR keyword selects elements from the list of containing elements or "ancestors" of the location source, counting upwards from the parent of the location source (which is ancestor number 1) to the root of the document instance (which is ancestor number -1).

For example, the location term

 ANCESTOR(1,*,N,1)(1,DIV) 

first chooses the smallest element properly containing the location source and having attribute N with value 1 and then the smallest DIV element properly containing it.

Note that the ANCESTOR keyword's second (element type) argument cannot be * or *CDATA.

6.3.7 The PRECEDING keyword

The PRECEDING keyword selects an element or pseudo-element from among those which precede the location source. The set of elements and pseudo-elements which may be selected is the set of all those in the entire document which occur or begin before the location source. (For purposes of the keywords PRECEDING and FOLLOWING, elements are interpreted as occurring where they start.) The result of the PRECEDING keyword is not guaranteed to be a subset of its location source.

The instance number in the location value of a preceding term designates the nth element or pseudo-element preceding the location source, counting from most recent to less recent. The XPointer

        ID(a23)PRECEDING(5,.)
      

thus designates the fifth element or pseudo-element before the element with an ID of a23. Negative instance numbers also designate preceding elements or pseudo-elements counting from the eldest to the youngest. The value ALL may be used to select the entire portion of the document preceding the beginning of the location source.

6.3.8 The PSIBLING keyword

The PSIBLING keyword selects an element or pseudo-element from among those which precede the location source within the same parent element. We speak of the elements and pseudo-elements contained by the same parent element as siblings; those which precede the location in the document are its elder siblings; those which follow it are its younger siblings.

The instance number in the location value of a PSIBLING term designates the nth elder sibling of the location source, counting from most recent to less recent. The location source must have at least as many elder siblings as the absolute value of the instance number; otherwise, the PSIBLING term fails.

 ID(a23)PSIBLING(1,.) 

thus designates the element immediately preceding the element with an ID of a23. Negative instance numbers also designate elder siblings, but counting from the eldest left sibling to the youngest. If the location source has at least one elder sibling, then the location term

 PSIBLING(-1,.) 

designates the very eldest sibling and is synonymous with

 ANCESTOR(1,.)CHILD(1,.) 

The value ALL may be used to select the entire range of elder siblings of an element:

 ID(a23)PSIBLING(ALL,.) 

thus designates the set of elements preceding the element with an ID of a23 and contained by the same parent.

6.3.9 The FOLLOWING keyword

The keyword FOLLOWING behaves like PRECEDING but selects from the portion of the document following the location source, not preceding it.

6.3.10 The FSIBLING keyword

The keyword FSIBLING behaves like PSIBLING but selects from the younger siblings of the location source, not the elder siblings. The XPointer

 ID(a23)FSIBLING(1,.) 

thus designates the element immediately following the element which has an ID of A23. Negative instance numbers designate younger siblings counting from the youngest sibling toward the location source. If the location source has at least one younger sibling, then the location term

 FSIBLING(-1,.) 

designates its youngest sibling.

6.4 String-Match Location Terms

A string-match sub-element address may optionally appear as the last in the series of location terms in an XPointer.

In this case the designated resource is a location which is found by searching the textual content of the current location source for occurrences of the SkipLit string given in the second argument. The Index is a number which selects among these occurrences, and the Offset is a number which gives a character offset from the start of the match to the designated location. Thus, the XPointer

 ROOT()STRING(3,"Thomas Pynchon",7) 

selects the letter P (seven from the start of the string) in the third occurrence of the string "Thomas Pynchon".

 ID(a27)STRING(5,'!',1) 

selects the character immediately following the fifth exclamation mark.

For purposes of string matching, the "text of the element" means all the character data in the element(s) in the current location source and descendant elements, all markup characters being ignored in the pattern matching. Thus in the example above, the string "Thomas Pynchon" would match and designate a reference in

 <authname><first>Thomas</first> <family>Pynchon</family>
        </authname> 

The pattern matching is exact and character-for-character. No case, space, or combining-character normalization of any kind is to be performed. Thus, there would be no match to "Thomas Pynchon" in the following:

        <example>thomas pynchon, 
        <auth><first>Thomas</first>
          <family>Pynchon</family></auth>,
        Thomas
        Pynchon</example>
      

7. Extended Link Groups

Hyperlinked documents are often best processed in groups rather than one at a time. If it is desired to highlight resources to advertise that traversal can be initiated, and if at the same time use is being made of out-of-line links, it may be an absolute requirement to read other documents to find these links and discover where the resources are.

In these cases, the Extended Link Group element may be used to store a list of links to other documents that together constitute an interlinked document group. Each such document is identified using the HREF attribute of an Extended Link Document element, which is a child element of the GROUP. The value of the HREF attribute is a locator, with the same interpretation as described above.

String-Match Term

These elements, just as with EXTENDED, SIMPLE, or LOCATOR elements, are recognized by the use of the XML-LINK attribute with the value GROUP or DOCUMENT.

Example 6contains sample declarations for the GROUP and DOCUMENT elements.

Example 6

<!ELEMENT GROUP (DOCUMENT*)> <!ATTLIST GROUP XML-LINK CDATA #FIXED
        "GROUP" STEPS CDATA #IMPLIED > <!ELEMENT DOCUMENT EMPTY> <!ATTLIST
        DOCUMENT XML-LINK CDATA #FIXED "DOCUMENT" HREF CDATA #REQUIRED >

The STEPS attribute may be used by an author to help deal with the situation where an Extended Link Group directs a processor to another document, which proves to contain an Extended Link Group of its own. Clearly, there is a potential here for infinite regress, and yet there are situations where processing several levels of Extended Link Groups is useful. The STEPS attribute should have a numeric value that serves as a hint from the author to any link processor as to how many steps of Extended Link Group processing should be undertaken. It does not have any normative effect.

For example, should a group of documents be organized with a single "hub" document containing all the out-of-line links, it might well make sense for each non-hub document to have an Extended Link Group containing only one reference to the hub document. In this case, the best value for STEPS would be 2.

Appendix A: Unfinished Work

A.1 Structured Titles

The simple title mechanism described in this draft is insufficiently flexible to cope with internationalization or the use of multimedia in link titles. A future version will provide a mechanism for the use of structured link titles.

A.2 Case Sensitivity in Attribute Values

It is possible to specify a link's resource based on the value of an attribute. It is is difficult to decide what the correct behavior is as regards case-sensitivity in matching. Ideally, the declared type of the attribute value should be taken into account, but that presupposes fetching and reading the document's DTD, which may not be appropriate in many XML applications. The current system, while easy to explain, may not prove suitable in the long run.

Appendix B: XPointers and Other
SGML Standards

Formally, the operations of the XPointer mechanism may be specified as operating on groves as defined in the HyTime standard. Every construct in such locators has a corresponding expression in SDQL, and most also have direct equivalents in the HyTime location module.

1. ISO (International Organization for Standardization). ISO/IEC 10744-1992 (E). Information technology--Hypermedia/Time-based Structuring Language (HyTime). [Geneva]: International Organization for Standardization, 1992. Extended Facilities Annex. [Geneva]: International Organization for Standardization, 1996. (See http://www.ornl.gov/sgml/wg8/hytime/html/is10744r.html).
2. IETF (Internet Engineering Task Force). RFC 1738: Uniform Resource Locators. 1991. (See http://www.w3.org/Addressing/rfc1738.txt).
3. IETF (Internet Engineering Task Force). RFC 1808: Relative Uniform Resource Locators. 1995. (See http://www.w3.org/Addressing/rfc1808.txt).
4. Sperberg-McQueen, C.M., and Lou Burnard, editors. Guidelines for Electronic Text Encoding and Interchange. Association for Computers and the Humanities (ACH), Association for Computational Linguistics (ACL), and Association for Literary and Linguistic Computing (ALLC). Chicago, Oxford: Text Encoding Initiative, 1994.

About the Authors

Tim Bray

321-3495 Cambie Street

Vancouver, B.C., Canada V5Z 4R3

tbray@textuality.com

Tim Bray is a Canadian. He entered the software profession in 1981; after on-the-job training from Digital and GTE, he became manager of the New Oxford English Dictionary Project at the University of Waterloo in 1986. He co-founded Open Text Corporation in 1989, and started an independent consulting practice under the name Textuality in 1996. He is a Seybold Fellow, editor of the Gilbane Report, and co-editor of the World Wide Web Consortium's "Extensible Markup Language Specification."

Steve J. DeRose

Chief Scientist

Inso Electronic Publishing Systems

1 Richmond Square

Providence, RI 02771

sderose@eps.inso.com

Steve DeRose work with hypermedia systems began with FRESS (at Brown University) in 1979. In 1989 he co-founded Electronic Book Technologies and designed DynaText, the first SGML online delivery engine, and other products. He is now Chief Scientist for Inso, EBT's parent company. He is active in standards with ISO, the TEI, SGML Open, and W3C, and is co-editor of XLL. He is a frequent speaker in industry and academe, and has written many papers and two books: Making Hypermedia Work (with David Durand), and The SGML FAQ Book.