Related Web Works

ECHT94 Trip Report

By: Mountaz Zizi Universit* de Paris-Sud Email: mountaz@lri.fr ---------------------------------------------------------------------------- The European Conference on Hypermedia Technology (ECHT94) was held in Edinburgh (Scotland) at the end of September this year.

In my opinion, there were three major themes at the conference:

* open systems and integration of heterogenous environments; * information structuring; * navigation and human-computer interaction issues with hypermedia systems.

Open systems and integration of heterogeneous environments

From the point of view of openness, the World-Wide Web is the most successful hypermedia system to date. Indeed, it is open to different information server protocols (Gopher, Wais, FTP, News, etc.), it is open to different types of nodes (bitmap, animation, structured text, etc.) and different types of links (static links, anchored links, computed links, etc.), and it has an open architecture that does not impose any constraint on the structure. There is no doubt that the simplicity of its basic data model, of the HTTP protocol and of the HTML format are the key to the success of WWW.

The amazingly large amount of data now accessible through long distance networks such as the Internet is another main reason for the success of the World-Wide Web. This also explains why other hypermedia systems tend to be more open than in the past. The experiment for providing Guide with network capabilities is one example. The experiment of C. Marshall at Xerox for providing flexible mechanisms to organize "at home" the information that comes from networks is another example. A last example is the project of Hill and Hall that consists of integrating that kind of information in Microcosm.

Another form of openness consists of integrating existing non-hypermedia applications in a hypermedia environment. The Universal Viewer from University of Southampton aims at this integration in the Microcosm environment. The Chimera project from Irvine has a similar goal for integrating heterogeneous programming environments.

Information structuring

Structuring information has been a fundamental issue in hypermedia from the very beginning. So, where are we now?

On the one hand, the basic model is getting more robust and more precise. This is partly due to the Dexter initiative which led to the definition of a unified abstract model that encompasses all specific models.

Dexter defines a three-layer architecture. The bottom layer is the "within-component layer" in which operations on the internal structure of entities are defined. On top of this layer, the "storage layer" defines operations on the overall network of nodes and links. Finally the "runtime layer" handles operations for presentation and user interaction.

Dexter also defines a data model in which the basic elements are components, composites, links and anchors. The concept of component is flexible enough to integrate external applications and to support dynamic contents. Indeed, so-called virtual components can be computed dynamically, e.g., as the result of a query. Composites are used to represent complex structures of basic components. Finally, anchors can be specified to reference a point inside a basic component without knowledge of its internal structure. This concept is used to keep the "within-component layer" and "storage layer" distinct. It also offers a way to interface these two layers.

The Dexter intitiative is not new, but it is still productive. Extensions of composites were proposed at the conference by Garzotto et al. and by Gronbaek.

Apart from the Dexter model, other projects aim at providing more flexibility for structuring information. Dolphin, for example, integrates several levels of structuring, from highly structured information such as typed nodes and links to unstructured information such as thumbnails or sketches.

Flexibility can also be achieved by providing the appropriate level of openness. Because of the diverse nature of data available on the net, open systems have to deal with large amounts of higly dynamic information. Information structuring cannot be done a priori, and it is never definitive because it changes constantly. Viki, from Xerox, proposes a technique for semi-automating the emergence of structure. Other systems such as Microcosm offer link services. In Shadocs, from LRI, links are created automatically based on the computation of similarities between nodes. Taleria has a similar approach. The Hart project, even though it has a slightly different objective (assisting the translation of paper documents into hyperdocuments), also provides techniques for semi-automating the construction of links and spotting potential anchors.

Navigation: presentation, interaction, querying

Navigation is not a new issue in hypermedia systems. Disorientation has been identified as a problem for a long time, since finding the most direct path to visit relevant nodes turns out to be more difficult than expected. Nevertheless, the conference showed significant progress in this area.

Concerning presentation issues, more precise output strategies were proposed. In CastingNet, from FujiXerox, information is presented interactively along axes. In Shadocs, Interactive Dynamic Maps are based on a metaphor of geographic maps with automatic layout in order to represent the information space and its semantics. Finally, Navigational View Builder from Georgia Institute of the Technology uses various clustering techniques to reduce the complexity of overview diagrams.

Concerning interaction models, various projects rely on the use of a spatial metaphor, e.g. Viki, and Shadocs. Kaplan also addresses this issue, from an ontological point of view.

Concerning queries, sophisticated models for querying hypermedia have been proposed by V. Chritophides and A. Rizk, and by B. Amman and M. Scholl.

Other issues

Apart from the three themes outlined above, other important issues were addressed at the conference.

Since the conference is about hypermedia systems, multimedia issues are an important concern. In this area, the extensions provided by Garzotto et al for Dexter composites support the representation of multimedia information and provide synchronisation mechanisms. Nevertheless, the way hypertext should face multimedia issues is far from being clear. While some argue that multimedia issues should be distinguished from hypertext and treated separately, others argue that hypermedia embraces both multimedia and hypertext issues that cannot be torn apart.

Historically, multi-user capabilities were one of the core issues in hypermedia systems (see the work on Augment from Doug Engelbart or the writings on Xanadu by Ted Nelson). Coordination problems that appear in WWW are just an example of such problems. Unfortunately, it seems that this area still has more questions than answers, and the conference did not bring any new findings.

Another unsolved issue concerns formats. Finding the most appropriate format for storing and exchanging hypermedia documents has always been an open problem. This year, the debate was about SGML and Hytime on one side, HTML on the other side, and more locally PDF (Adobe's format). The panel entitled "HTML: poison or panacea" was the opportunity for purists to point out aspects of HTML that do not fulfill SGML requirements. It was also the opportunity for HTML advocates to argue that SGML or Hytime are still too complex for most occasional users whereas HTML is a widespread operational format. This is particularly important for hypermedia systems such as WWW where occasional users can also be authors.

Conclusion

The conference has shown that hypertext and hypermedia is an active and mature field. After a first generation of mostly static systems, we are now experiencing a second generation of hypermedia systems that feature more flexible structures and specific mecanisms to support dynamic information and openness. Undoubtedly, the success of systems such as WWW demonstrates that hypermedia will eventually play a key role in the future of information systems. ---------------------------------------------------------------------------- Keith Instone / instone@acm.org / 95-01-03