XML Developers' Conference

August 20-21, 1998

Montréal, Canada

Presentation Abstracts

    David Brownell
    Java Software Division
    Sun Microsystems, Inc.

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/stevew.htm]

XXX - eXpandable XML eXploitation

By Steve Withall

This presentation describes a number of design ideas for flexible, expandable applications that manipulate and otherwise exploit XML documents. Steve will demonstrate these ideas using a Java 'XML Testbed' application. (Steve intends to publicly release all his source code either at or prior to the conference - including his own parser and XSL engine.)

Steve divides the application into four parts:

Four Java packages provide functions for representing and manipulating XML documents:

• XML Atoms (package xa) to encapsulate low-level knowledge of the XML language itself. While normally embedded in parsers themselves, the idea is to allow other software (eg. an editor) to use this information.
• Grove (xg). Nothing revolutionary here!
• XML Manipulator (xm) provides a range of document manipulation capabilities, including factory classes, a base parsing 'engine' class, and document model and document listener classes (to co-ordinate changes made in multiple document views). This package also contains a registry for specialised element and other classes.
• Customizer (xc), fulfilling the same role as JavaBean customizers via a collection of classes to display and edit the standard types of XML entities.

Any supported parser can drive the application, in a similar manner to SAX, but at a higher level. (SAX is, however, a supported 'parser').

The vanilla application manipulates vanilla XML documents, but semantic knowledge can be embedded in specialised classes. Steve will demonstrate three such usages: an XSL engine, UI configuration, and rendering a database schema in XML via JDBC. The basis for expandability is to 'register' information about specific XML element types:

• Element class name. Specialised element classes can contain whatever specialised processing they wish. Steve (sacrilegiously) invokes a 'verify()' method on each element when its parsing is complete, to perform specialised checking during parsing.
• Customizer class name, for specialised display and editing.
• Parser class name, primarily to allow special PI parsers to be slotted in.

Steve will demonstrate these ideas using his XML Testbed application, which implements them all and is also a basic XML authoring environment. Steve hopes to demonstrate nearly-WYSIWYG XML editing driven by XSL style rules - if it's working in time.

Steve Withall has been developing commercial software for eighteen years, the last eight designing large-scale object-oriented systems. He currently works for Access Systems in Sydney, a company formed to develop Interactive Television applications. For this, he and his colleagues developed a language for describing ITV user interfaces: PNML - Pretentiously Named Markup Language. (Why couldn't XML have been around then?!) While the company has moved on from ITV, PNML is still used for various purposes. Steve comes to XML from a very different direction than the SGML gurus!

Responsible for, among other things, extracting documentation from thirty unruly software engineers, Steve has an active interest in automatically generating system documentation where possible, and is using XML and XSL to develop solutions for this problem.

Note: The details of the Java 'XML Testbed' application used to demonstrate these ideas are now documented online, and the software is available from the W3C web server. Slides from the Montréal presentation are also available. - rcc, 980827

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/holman.htm]

XML Conformance

By Ken Holman

[This would be a brief report.]

At the first XML Developers' Day in 1997, OASIS (the Organization for the Advancement of Structured Information Standards) announced the early formation of a technical subcommittee to explore XML Conformance. Over the first year, this committee has explored the requirements for XML Conformance while the XML Community has finalized the standard and begun real-world implementations of the technology. This presentation briefly overviews the status of the work of the subcommittee, the approaches being considered for use in the committee results, and the needs for input from the XML Community that will help advance the effort.

Note: The presentation slides for "Report on XML Conformance" are available in HTML format. See the main database entry, XML Conformance.

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/brian.htm]

Filtering XML Documents and Data Sets

Brian McFadden, Xtenit, Inc.

To meet the needs for several content management applications, we are developing a general filter that will use a simple yet powerful syntax for selecting and filtering XML documents and data sets. This filter will be used in several applications for routing, categorizing , and retrieving XML documents and data sets. Characteristics and advantages of the syntax used are:

• Standard if-then-else form with boolean Operators (AND, OR, NOT)
• Support for standard operators (<, >, =, etc.) for strings and numeric data
• Scope specific (i.e. (Parent).(child).(name) ) and global (i.e. *.(name) ) element selection
• Implicit support for elements with multiple occurrences, object data, and tabular data
• Able to reference all elements and attributes in any well formed XML document.

The core process for the filter supports distributed sets of rules, and each rule can reference data from multiple documents and data sets, (for example document selection rules could reference both profile and content).

Examples of applications include:

• Information Routing - Intelligent routing applications match XML content to profiles (also in XML) based on rule sets that contain both general rules and rules specific to the each content and profile item. This allows delivery of both general and specific real-time, event-driven information to the appropriate audience.
• Refined Text Retrieval - The filter is used to assist in indexing of stored documents. Applications sitting on top of the indexed archive use the filter and rule syntax to refine searches and eliminate extraneous results.
• Agent Servers -- Agents defined by a profile (in XML) and a set of rules use the filter to process and categorize a stream of XML content.

Our presentation will discuss our approach, summarize the syntax used, and give specific examples of applications with XML documents and DTD's.

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/papresco.htm]

Python: the 100% Buzzword Compliant XML Processing Language

By Paul Prescod (ISOGEN International Corporation)

Python is a simple, powerful programming language. It is designed to solve many of the same problems addressed by Perl (text processing), TCL (embedding), Visual Basic (user interfaces and prototyping) and to a lesser extent, Java (distributed, portable applications). Instead of being yet another over-specialized "scripting" language, Python shares its best features with many great languages and integrates them in a predictable way. With Smalltalk and Scheme it shares flexibility and simplicity, with the Algol family, a familiar infix syntax, with the Simula family, object oriented programming features and with Perl, an open development model, powerful text processing features and a versatile set of standard libraries (including XML processing tools).

These features make it perfect for processing XML and SGML documents efficiently and conveniently. The Python XML Special Interest Group has been charged with the task of making libraries and documentation for processing XML and SGML documents, W3C DOM objects, ISO Groves and SAX data streams.

Python code can run in a highly portable, Python-optimized bytecode interpreter, or in any Java virtual machine. When run in the Java environment, Python code can be distributed across the network in JVM classfiles. This code can use any Java classes and libraries (including SAX parsers and the AWT). Thus, Python can be used either instead of, or in addition to Java, especially where a project requires faster coding or more flexibility than Java provides.

In short, Python is an easy to learn, scriptable, portable, distributed, multithreaded, web-aware, embeddable object oriented language for everything from rapid prototyping to large system development. Thanks to the efforts of the XML SIG, it is also shaping up to be a very robust platform for XML processing. This talk will describe the products and plans of the SIG.

The presentation slides are available online.

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/tbray2.htm]

Reifying the DPH

    Tim Bray
    Textuality

Meeting the needs of the symbolic DPH (Desperate Perl Hacker) has always been one of the key design goals of the XML project. Fortunately, the perl community has decided to meet us half-way. This short presentation describes the work accomplished to date on giving perl the ability to process XML, which has involved wiring in the expat XML processor.

The work has progressed far enough to allow the construction of a variety of different experimental interfaces to XML documents for perl programmers. Several such APIs are now available. I will describe how one goes about constructing such APIs, and will demonstrate the use of one aimed directly at the DPH.

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/rbrown.htm]

InDelv XML/XSL editor and Java class library

First public presentation of a wysiwyg XML editor which accurately renders XML using XSL style sheets. This software was developed by rewriting a commercial text class library from scratch, specifically for Java and the new XML family of standards. It creates documents which can be displayed or printed internally, exported as XML / XSL for use in other tools, or converted to HTML for viewing by traditional web browsers.

XSL is used as the internal rendering model with DSSSL based flow objects and format characteristics. Support for the DSSSL formatting model roughly corresponds to the DSSSL Online proposal of August 96. XSL will be supported to the greatest extent possible relative to the XSL Working Draft due in July 98 (hopefully including element reordering, literal text and named styles).

The document model implements DOM (Core) Level 1 and parsing is handled by any SAX compliant XML parser. The linking and text selection model is based on XLL. Key design goals include fast on-screen rendering, accurate printing and flexible application setup. XML is used to specify menu definitions, key maps, view and window layout, event hooks, nls string translation and user preferences. The intent is that 'XML applications' can be created and deployed by non-programmers and so far the results have been promising.

The version being presented is a 'Technology Introduction' and is intended for use by experienced programmers and visionary XML hackers. Some capabilities will be limited. In particular, it will have minimal support for DTD's and traditional SGML authoring features. The editor can enforce hierarchical integrity and infer types for new elements, however much of the work planned in this area will wait for DOM DTD support to stabilize and a validating SAX to emerge.

A retail product based on this class library is targeted to ship in late August (Java and Smalltalk) on a fully supported basis. It would be useful for developers who want to begin working visually with XML now and yet are willing to wait for additional features. More details will be made available at the session.

Note: InDelv has now annnounced its plans to offer a new commercial XML/XSL editor in mid-September. It is to support the new XSL working draft released by W3C on August 18, 1998. [announcement, local archive copy] -rcc 980826

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/tbray.htm]

Construction of the Annotated XML Specificiation

Tim Bray

The idea of the annotated XML specification is simple; a collection of relatively small pieces of commentary attached directly to the electronic version. Since there are some parts of the XML spec that clearly cause people trouble, it is important that the annotations be very highly localized.

The decision to use XLink and XPointer was easy, and the result was gratifying, but the hurdles along the way were of considerable technical interest. The included:

1. Implementation of partial XLink and XPointer processors.
2. Design of a language to contain the annotations and all the XLink and XPointer apparatus.
3. Bridging the gap between the XML form of the spec and annotations, and the XML-oblivious browsers of the world.

This presentation will introduce each of these technical hurdles, describe the chosen solutions, and offer a hindsight-based opinion as to what the correct solutions might be.

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/peterm-r.htm]

The Virtual HyperGlossary - adding semantics and ontology to XML

Peter Murray-Rust and Lesley West

VHG Consulting, Ltd

XML allows authors to create their own markup (e.g. <DOMAIN>), but no direct handle for answering 'what does this mean?'. One of the most powerful (but very cost-effective) ways of adding meaning is by linking to terminological resources such as glossaries or dictionaries. Although there are a large number of HTML-based glossaries (often of very high quality) on the WWW there is no consistent syntax. It is almost impossible to search for terms in a robust and meaningful manner. For example the semantics of brackets in "Domain (DNS)" are different from the semantics in "eXtensible Markup Language (XML)". The first is a qualifier, perhaps distingusing the term from "Domain (protein)", while the second is an acronym.

We have developed a simple but scalable DTD for terminology based on ISO 12620 (Data Categories for Terminology). This DTD uses a deliberately small subset of about 12 categories (e.g. <term>, <acronym>, <synonym>, <abbreviation>, <definition>). Others can be added through an attribute-based syntax. Hyperglossary is used in a wide sense to include any semantic resource composed of standardised subcomponents such as data sheets or catalogues. Thus a molecular hyperglossary could include chemical structures, measured properties and commercial availability.

Because XML is tree-based it supports hierarchical collections (thesauri, catalogs, etc.) in a natural and powerful manner and we have found that most of our current examples fall into it. For non-hierarchical relationships (<see>, <seeAlso>, multiple broaderTerms, etc.) the VHG uses the full power of XLink to add additional structure. Thus terms can be grouped in different classifications by using xml:link="extended" with locator references to the linked terms. Equivalences (e.g. in multilingual glossaries) can be defined through an external link database. This allows different language curators to develop their glossaries in parallel and link the terms through XML ID's.

Curators are a key aspect of VHGs and we expect that many VHGs will be XLinked to 'neighbouring glossaries' especially where the curators know each other. This avoids the pitfall of a 'grand universal classification' and allows readers to explore multiple interpretations of the same term. Learned societies and other authoritative providers of public information can benefit by making their current terminologies (often paper-based) available. We shall show an example from a collaborating International Union and we are also partnering with providers of commercial semantic resources.

For organisations developing 'knowledge capture' strategies, terminology is an essential component. By encouraging distributed hyperglossaries on an Intranet an organisation can empower invidividual units to capture and manage their terminology without excessive central overheads.

The VHG is applicable to any domain, but has specific support for technical subjects. We have encoded scientific units (SI and non-SI) into VHG format and this means that unit conversions (currently very problematic) can become automatic.

The VHG DTD is completely implemented in JUMBO-VHG (JUMBO is a freeware system based on SwingSet) and demonstrations of many applications will be given. These will include component-based creation and editing of glossaries. We shall also show Virtual HyperMarkup which is the detection of terms in XML-based documents and their linking to appropriate glossaries. This enhances the semantic value of the document and should also lead to normalisation of information if glossaries are widely available, e.g. through cascading glossary servers.

Example

References

(a) Also at Virtual School of Molecular Sciences, University of Nottingham, UK

(b) VHG is being registered as a trademark by VHG Consulting, Ltd.

Note: See the main database entry, "Virtual Hyperglossary (VHG)."

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/fensterm.htm]

Using XML in Knowledge Management

Kurt D. Fenstermacher

We are building a system (called SingleSource) for information capture and access within a strategic consulting firm. Within the firm, consultants working with clients request information on topics ranging from: "What's the GDP per capita of India?" to "What factors will impact the demand for petroleum products over the next decade?" While researching these questions, information professionals need to search and collate results from many sources, both online and offline. Our prototype provides a front-end for researchers within the organization to search multiple online sources, view clustered search results and finally produce a finished document summarizing the requested information.

Smart information retrieval

The SingleSource environment will aid researchers in finding information, while simultaneously storing information about their search strategies. As the system "learns" search strategies for kinds of search queries, it will use that information to improve later retrievals. For example, if one researcher (who specializes in the automotive industry) frequently uses only a handful of sources to answer questions, the system will store an association between automotive questions and those few (but highly relevant sources). When a less experienced researcher has a question about the automotive industry, SingleSource will focus the queries on the best automotive sources, and recommend the automotive specialist as a contact.

XML in SingleSource

The prototype system depends on XML (or XML-based technologies) in several ways (each description is followed by a note regarding its likely stage of completion by conference time):

• Many system components often need to share information, and XML documents are used to pass object-oriented information from component to another. For example, a single search query is sent to multiple information sources, and the results are combined and then clustered into semantically similar groups. The clusters (which can themselves contain clusters) are represented in a tree structure, which is then output as an XML document. (Will be completed by conference.)
• Knowledge capture is an important aspect of SingleSource. In particular, we plan to store information about searches as they are conducted. (For example, we will track which sources were used for a query, and which sources consistently return the most useful results over time.)  We are currently working on a process to generate this meta-information in an XML-based format (such as RDF or MCF). (Will be completed by conference.)
• To be a useful front-end, the system must seamlessly integrate many different information sources (including Web-based, intranet and CD-ROM databases, for example). Representations of what data is available, and how to access it, must be built for current sources. In addition, new sources should be easily described, so they can be incorporated into SingleSource on-the-fly. We are currently evaluating WIDL for this purpose, although a custom-designed alternative would also be XML-based. (Will be completed by conference.)
• SingleSource generates extensive information about the context of a task (which will be stored in an XML document) and that task information must be stored and managed persistently. Because we can easily transfer data between XML and object representations, we are incorporating an object-oriented database (Object Design's ObjectStore) to store XML and other object data. (Likely to be completed by conference.)

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/kimber.htm]

Eliot Kimber Presents: Grove-based XLink Implementations

Eliot Kimber (Isogen) will give a demonstration of PHyLIS, Personal HyTime Link Information System, an attempt to implement the HyTime standard using componentized software techniques (ActiveX). PHyLIS uses a literal grove-based approach and demonstrates the ability to apply XLink, SGML architecture (AFDR), and HyTime processing to XML documents in a completely standards-based environment. PHyLIS is designed to be an open integration platform that is infinitely extensible. See www.phylis.com for details, including source code.

Eliot will also demonstrate an implementation of an XLink processor developed by Masatomo Goto of Fujitsu Laboratories. This processor uses an implementation of the XLink property set based on the HyTime property set. Supported features include:

• Simple and Extended links
• Unidirectional and multidirectional links
• Inline and out-of-line links
• Extended link groups
• Attribute remapping (xml:attributes)

Fujitsu Laboratories plans to integrate this XLink processor with its "HyBrick" browser later this year. For further information on "HyBrick", see http://collie.fujitsu.com/HyBrick.

Note: Presentation slides for "Grove-based XLink Implementations" ['Xlink and Groves'] are available online. See also the main section for XLL (XLink, XPointer).

[GCA Website URL: http://www.gca.org/conf/meta98/xmldev98/megginson.htm]

XAF: Using Architectural Forms with XML

David Megginson, Megginson Technologies Ltd.

Architectural Forms are an International Standard (ISO/IEC 10744:1997, clause A.3) for applying multiple inheritance to XML and SGML documents.

Megginson Technologies has recently released the first version of XAF, a general-purpose Architectural Forms processor for XML using the widely-supported SAX (Simple API for XML) interface. XAF is available free of charge at the following location:

http://www.megginson.com/XAF/

This presentation begins with a brief overview of Architectural Forms, then compares Architectural Forms with XML Namespaces to show how the two specifications can work together.

The second part of the presentation introduces the XAF software itself and examines the design decisions and compromises that were necessary to implement Architectural Forms using XML and SAX.

Note: David Megginson's presentation slides are available online in HTML and PDF format. For related information, see the main section: Architectural Forms and SGML/XML Architectures.