[This local archive copy mirrored from the canonical site: http://www.texcel.no/se97talk.htm; links may not have complete integrity, so use the canonical document at this URL if possible.]

Why Your Document Management System Should Care About Hyperlinks

Or, why you should care about a document management system if you care about hyperlinks

Presentation by Paula Angerstein

of Texcel Research, Inc.

at SGML/XML '97

on December 9, 1997

Why links

A link is some construct to represent a relationship between two or more things
Historical use is straightforward association of two data items, e.g., a cross-reference
Historical venue for links is hypermedia systems
Emerging use of links
- to locate distributed objects
- to specify dependencies
- to associate “metadata” with data

Some common link requirements

Links to arbitrary data and to points within that data
Links within and across documents
Links with multiple endpoints
Links that carry with them some set of semantics, such as a type and behavior
Links into data that cannot be modified
Control over the direction of traversal of a link
Control over what types of objects a link can point to
Notification when a link end is invalidated or modified
Version history of a link
Context-sensitive links

Representing links

SGML ID/IDREF
HTML
HyTime
XML Link

Well-accepted properties of links

Specification of the link itself usually via some combination of elements and/or attributes (link recognition)
Specification of how to find the endpoints (addressing)
What the link is for (role)
What to do when the link is “activated” (behavior)
Allowed types of things the link can point to
Allowed direction of traversal between link endpoints
Various other metadata about the link itself: a descriptor, who created it and when, system-specific instructions, etc.

Link lifecycle

Typical user uses links: hypermedia systems present and traverse existing links
Somebody has to create and manage links: authoring and document management systems must do interesting things with links
Linking within constantly modified data presents some hard problems:
- Addressing that works when linked data is modified
- Ongoing validation that links are still valid
- Automated synthesis of links

Why ID/IDREF doesn't cut it

Limited to one document
Limited to addressing an element
Not enough standard semantics
Can't maintain links independently of the data
Impossible if read-only data doesn't already have IDs

Fundamentals of HTML links

The “A” tag is a link
It has an “href” attribute whose value is a URL
That's it

This is an HTML link: <a "href=http://www.texcel.no/texcel.htm">

HTML link shortcomings

No links to spans of text and spans of content
No links into arbitrary data types
No links with multiple endpoints
No links independent of data
No control over the types of endpoints of a link
No control over the direction of traversal of a link

Fundamentals of HyTime links

Hypermedia/Time-based Structuring Language, defined in ISO standard 10744
In the standard, HyTime defines architectural forms: a set of “meta” element classes and attributes with standard semantics
When you write a DTD, you make an element a link by applying an architectural form via a “HyTime” attribute
End result is that instances of the element in a document are links
A link relates two or more link ends. Each link end is a locator to a piece of data known as an anchor.
A link end addresses the anchor via various mechanisms such as nameloc, treeloc, queryloc, and dataloc.
A contextual link gets one of its link ends from the link element's position in the document.
An independent link resides independently of any of its link ends.

This is a HyTime link: <clink hytime="clink" linkend="TexcelLogo">

Shortcomings of HyTime

According to some people, it is not possible for HyTime to have shortcomings because it can do anything
This is its shortcoming

Fundamentals of XML links

Any element becomes an XML Link when it has an attribute named “xml-link”
The “href” attribute is the locator and is a URL
Additionally standardizes the fragment id and query portions of a URL as either ID referencing or TEI Extended Pointers (XPointer)
Other attributes specify information about the link such as its role and behavior
Simple links are like HyTime contextual links; extended links are like HyTime independent links

This is an XML link: <simple xml-link="simple" href="file:///C|/texcel/im/lib/texcel.gif">

Shortcomings of XML Links

No links into arbitrary data types
No control over the types of endpoints of a link
No control over the direction of traversal of a link

Links “inside” an SGML document management system

SGML document management systems typically have unique object identification for every SGML element
These repository identifiers (RIDs) make complex addressing unnecessary: link resolution is simple
Within its own domain, a system can provide efficient link storage and manipulation
When data goes out of this domain, links can be exported to a standard form

Link creation

Present candidates for link targets, e.g., via tree views, query results, views of content
Generate an address to a link target
Automatically generate ID values
Ensure links are only to allowed types
Associate link type and other information with the link
Update an independent link map
Automatically create links

Link maintenance

Integrate with authoring systems to prevent deletion of link targets
Notify when link target contents are modified
Notify when an address locates a different target
Potentially recalculate addresses
Retrieve and update link metadata
Maintain an independent link map
Maintain context of link applicability
Trace link lifecycle

Link delivery

Real-time link traversal
Determine and export a web of linked data
Export links in a form optimized for the delivery system
Drive conversion and delivery processes

There's more to links than viewing

Tangible benefits of planning for links across the entire document lifecycle
Exploit the capabilities of your SGML document management system to support linking
Quality gains are certain to follow