This was the third US-based hypertext conference (following on from the '87 and '89 conferences) with 575 participants, 450 from the USA and 125 from the rest of the world.
In general I found attending the conference not as interesting as the ECHT '90 conference held in Versailles last year. I think the main reason for this was the way the demos were handled. These are an important way of meeting groups currently active in the field. After attending the "intellectual" sessions all day the demos were held in the evening. By that time I had little energy left to see and understand the demos, let alone try to conduct an intelligent conversation about them.
I also had the feeling that most of what I heard I had heard before. This was probably because I mostly attended the Technical Briefings and Panel sessions rather than the Paper sessions, and now that I have looked through the papers they do make a good collection.
One of the highlights of the conference was the San Antonio Information Kiosk developed by Elroy D'Souza, Tammy Duncan and John Leggett from Texas A&M university. It contained information about the conference events and contributors, and suggestions for places to visit in San Antonio, including maps and photos. This was written using IBM's Audio Visual Connection running on a PS/2 Model 80. The authors had gone to extreme lengths to collect attractive and relevant material to put into the system, and had taken care in designing and evaluating the interface. They had done a great job (with little help from the authoring system they were using).
This document is a summary of what I learned from the conference and is intended to give a quick means of access to the papers in the proceedings, as well as an insight into what went on at the conference itself.
The main theme of Pat's talk was that people need appropriate tools for the job. Whether you are evaluating interfaces in general or using hypertexts in particular it will help if you can call upon appropriate tools. During the talk the word "tools" gradually slid via "support" into "prostheses" and emphasis was given to the need to offset the cogitive demands of working with electronic documents by providing support for cognitive activities. The existence of these cognitive demands was illustrated by the results of a number of laboratory studies, some showing that for inexperienced computer users even a click could have cognitive costs.
Hypertexts not only have to be designed and created with the user's task in mind, but evaluation has to be carried out with other factors in mind:
· adequacy of the content and the interface,
· acceptability which may depend on the alternatives readers know about,
· adaptability by readers for the task in hand,
· skills of the readers as information users,
· costs of production and dissemination.
Pat challenged the notion that a criterion of design adequacy was met as long as the information was there if readers wanted it. In a small study of people reading text, where readers had a free choice about when to access a diagram, they chose not to access it. However, when the hypertext was constructed so the readers came across the diagram during their reading, they built up a better understanding of the material. Thus readers themselves do not necessarily know how to best exploit the hypertext but hypertext designers can compensate for some of these skill deficits.
Readers' strategies were further explored in tasks where readers could access glossary information by clicking on unknown words in the text. Again it was found that readers' willingness to access this information varied with the interface characteristics. The message:that design matters.
Pat cautioned against a too simplistic approach to evaluation. Fastest might not always be best. In one study people were given two ways of navigating round a hospital, one group used only the existing notices and wall signs, the other also had a map. People with the map found it helpful, but those without the map were faster. Nevertheless, time spent studying the map and making an informed decison about how to reach the target destination was considered time well spent by those having the maps.
A well designed hypertext will take into account how people work with information (e.g. how they search, filter, find, understand, decide, plan, remember, manipulate). Computers can offer many kinds of support for these activities. In order to illustrate this, a study of the use of memory aids was reported. Users were given different ways of navigating through information and had access to different tools to help them remember what they were looking for and what they found. The use of these aids varied with the difficulty of the current task. Thus users were making strategic decisions about which aids to use and when. Usage also varied with the characteristics of the interface. This indicated that some design decisions had cognitive costs for users. It also suggested that monitoring the use of assistance such as memory aids might be a valuable addition to the evaluator's tool kit.
One incidental observation came from pilot studies of how to explain the interface and the tools to those taking part in these studies. It was found that if the tasks being done became difficult before people knew about the various memory aids, then they welcomed their arrival and quickly learned how to use them. It may be an important aspect of cognitive prostheses in hypertexts - that readers need to see them as contributing to the solution of their problems rather than as being yet another problem in their own right.
Not only was the talk itself entertaining and enjoyable, but it, and the version in the proceedings, contain results which are extremely relevant to the way we should approach supplying tools for readers in an environment as complex as hypertext.
Unfortunately the talk is not written up in the proceedings, but participants did receive copies of the slides that were used.
At the end of Halasz's talk during the 1987 conference he saw the future of hypertext as its own obsolescence. However, the world of hypertext has broadened to include:
· date storage (persistent, shared multi-media objects),
· data models (networks, hierarchies, sets, relations etc.),
· navigational facilities (browsers, full-text search, guided tours etc.),
· applications (authoring, knowledge representation, interactive fiction, CASE etc.),
· issues of situated use (hypertext engineering, rhetorics, evaluation, disorientation, copyright laws).
There will be plenty of challenges for the years to come.
We can impose four dimensions on the field:
the navigators vs the architects
the navigators focus on the nodes and their contents, and navigate from node to node in the information space (KMS, Hyperties, [Guide]), the architects focus on the overall structure of the hypertext and manipulate it using a browser or an overview (gIBIS, Aquanet);
the literalists vs the virtualists
the literalists create links explicitly, and navigation is by following these links (Intermedia, NoteCards, KMS, Aquanet, [Guide]), the virtualists see the structure as implicit in the contents of the nodes and the system can compute structures over the nodes for the user (SuperBook, CYBERMAP);
the card sharks vs the holy scrollers
the card sharks view each node as a fixed sized "card" onto which information can be placed, navigation jumps you from card to card (KMS, HyperCard), the holy scrollers use a collection of lengthy documents and as well as jumping between documents users can scroll within a document (Augment, FRESS, Superbook, [Guide]);
the literati vs the engineers
the literati see hypertext as the next new medium leading to a new definition of writing, while the engineers see it is an information tool for enhancing ways of finding information.
Searching can be carried out on content and on structure, though the former has had more work done on it than the latter.
SuperBook is a hypertext system with no links, and navigation is done using full-text search. Links aren't essential for hypertext, but what is a link anyway? Aren't queries just "dynamic" links?
Models for hypermedia now include composites (groups of nodes and links handled as discrete entities), e.g. Dexter, but why are we limited to the node/link model. Petri nets, sets and relations are possible alternatives.
There has been little work done on virtual structures [e.g. "link to the most recent mention of cherry tree"], but some on related concepts, such as:implicit links (from a word to its dictionary entry); contextual links (links that appear or disappear depending on the state of the system or the user); discovered structures (links or composites created based on similarities between contents of nodes); computational nodes (when accessed show the result of an external computation). We could go further and have the entire structure of the space created by the system.
Work has been done in this area, largely for knowledge representation, although the computation issue is separate from the knowledge representation issue. [Sorry - I'm not clear what he means here.]
"A good versioning system will allow users to maintain and manipulate a history of changes to their network. ... simultaneously explore alternative configurations." However, nothing has been published (since CSCW '86), but maybe this is because it's not a critical requirement, for example it's not mentioned in the "Industrial Strength Hypermedia" paper (see p 13 of this report), or maybe because we're so used to our operating systems we don't know what we're missing.
There are two things that need to be supported - the mechanics of multi-user access and social interaction. Multi-user backends are becoming standard, but there are few hypermedia specific facilities as yet. Rhetorics for hypermedia is an issue on its own, but nothing is being done on mutual intelligibility. Systems for supporting specific collaborative processes have been created (Aquanet, InterNote, gIBIS, PREP).
Some systems have facilities for tailoring or extending the system, e.g. scripting languages, schemas and templates, paths and guided tours.
The research agenda needs to be revised. [There was a diagram showing how the previous 7 categories have been regrouped into 7 new technology categories, and 3 marketing issues have been added. I have indicated this by each issue.]
(Was issues 1, 2 and 3.) We need to include non-network structures and virtual structures in our notion of hypermedia. A broader array of "data-models" will provide opportunities for integration with complementary technologies.
What is the appropriate decomposition of hypermedia functionality into independent components? What are the appropriate communications protocols for coordinating these components? How can we take advantage of the rapidly evolving standardised mechanisms for (distributed) object-oriented inter-application integration?
Interfaces that allow users to manipulate large network structures is a long-standing problem. Many partial solutions have been proposed:fish-eye views, 3D information visualiser, nested set displays. Aquanet can currently be used to browse more than 1500 objects.
The challenge is to build a usable industrial strength hypertext system capable of handling 10,000 documents (with 100 word processor formats).
Why isn't hypertext selling like hotcakes? As a field we are too technology driven and do not know our markets. Who are our customers and what do they need? (See "Industrial Strength Hypertext" paper on p 13 of this report.) There is a broad range of choices - technical manuals, interactive fiction, CASE. We need to develop an understanding of the costs and benefits for these applications.
Users are needed to drive this user-centred technology, and, more importantly, as the market goes so goes the research funding.
We need to develop and start using standards, in areas such as:anchoring in desktop applications, hyperdocument interchange (HyTime), link services, document/anchor addressing in LAN' and WAN's. Hypermedia standards need to complement related standards, and be able to progress with the field.
The technical briefings were a new form of presentation introduced to the conference this time. "Technical briefings provide a presentation medium for presenting details of a concrete design rather than an empirical or theoretical contribution. It is intended for designers to communicate valuable insights and experience to implementors and designers. It should be a frank discussion of the decision points and trade-offs involved in the design of some hypertext system or application."
I think this is an excellent idea for the hypermedia conference, in that there are many designers who can learn from others' experience, where a formal scientific paper is not a good enough medium (for showing the interface for example). The results were mixed. Too often only "what" was described, instead of describing the design possibilities, the reason for choosing one design over other possibilities, and the experience gained (positive or negative) in the use of that design.
Meyrowitz and a colleague from GO Corporation gave a demonstration of the PenPoint system running on a notebook computer. (This was interesting, though was more a demonstration than a discussion of the design issues.)
The pen-based computers will be produced by different hardware manufacturers and have no moving parts. They have a large screen, internal memory and the main means of input is a pen (a keyboard can be plugged in).
The PenPoint operating system uses the notebook as a metaphor for organising the files in the system, and is based on object oriented technology. Material is mainly input as postscript, but GO want to keep things flexible.
Gestures with the Pen are used as commands to the system. For example, a carat gesture (^) means insert, a cross (X) means delete, circling words means search for this pattern. Separately written letters can be recognised (after training).
As Meyrowitz pointed out, electronic books have to be better than paper books, and an electronic book should be a reading and a writing device, and all applications should have hypertext functionality.
One marketing question is, who is currently using keyboards and who will use pens?
The talk was more about creating documents in hypertext rather than the design decisions taken in building the hypertext system.
The paper in the proceedings describes the Storyspace system, used as a writing environment.
WALT is an implementation of the Andrew toolkit, which required 8 person-months of effort to create. It runs under XWindows.
They have partially implemented a hypertext based on the Dexter model, which will allow them to incorporate document structure features into advanced information retrieval algorithms.
The screen has seven components, which can be resized, but their relative positions are static, so each user has a uniform interface. A scroll bar is used instead of multiple windows.
The information is divided into books, which are placed on a book shelf. Each book can be selected to give a book spine with a representation of the chapters. The current name of the book and chapter is displayed on the screen (top right in picture on p 391). A book has different levels, and a table of contents.
Information can be included in patients' records (the system is used in a medical environment).
When users do searches they are asked if the section they were taken to was relevant or irrelevant. This information is then used to give a best approximation for the current search. The number of hits are displayed on the spines of the books.
Two different representations are used for the chapters in a book, depending on how many there are. For small numbers a bar chart is used to show for each chapter how many hits there were from the search. For larger numbers the spine is divided into a checkerboard, and each small rectangle (representing one chapter) is given a shading from white to black, indicating the density of hits.
The system is used in the collaborative work of scientific groups. It supports links between different types of media. Links can be created across sites. The system uses a distributed architecture. Everything within the system is treated as an object.
The goal of creating the Hypertext Compendium was to explore hypertext publishing, first of all by building a hypertext database with a significant proportion of the material in the field, and then publishing that in multiple formats.
The compendium currently contains a few conference proceedings and is available in KMS, HyperCard and ASCII versions.
One of the things that were learned during this project was that the index was difficult to create, first in choosing the words to be in the index and then deciding which occurrences should be included in the index. There is also a distinction between linking to a whole article or linking to part of an article.
The notion of stepping-stone frame was created, where when the user chooses a word to ask for information, rather than going directly to a destination (be it the index or a related item elsewhere in the collection), a list of possibilities is given. For example with a standard reference, a list of where else this article is referenced is presented.
Akscyn also gave guidelines for editors. Don't include links in other people's work without asking, with exceptions for citations, "see also", and references to figures. Build many small tools to help with the process, especially for checking things. However, don't build the tools too soon. Explore the design manually before investing the effort in automating it. Develop a house style by using a small subset of the material.
Akscyn would like to see the project grow and develop with a federal editorship model, so that many people can contribute, and the material remains high-quality. Reviewers and editors are instrumental in this process.
Unfortunately I missed this presentation. There is a write-up in the proceedings though. "In addition to allowing users to create and send links, InterMail automatically generates links between messages. ... messages constituting a conversation between users are chained together with automatically generated links."
The proceedings contain a full write-up of the presentation. Their system comprises a back-end relational database and a hypertextual front-end. "The back-end is designed to address the needs of database administrators... . The front-end is published and distributed periodically via floppy disk to users, who need a fast, easy-to-install and easy-to-use read-only environment for browsing and searching information."
Panels are always hit and miss - you can sit bored for an hour then someone comes up with an inspired remark. I haven't given my overall opinion of the individual panels, but pulled out the pieces I found interesting.
I can only give a biased opinion of this panel, since I was one of the speakers. Our intention was to show four different commercially available hypertexts and give our own review of them in terms of how suited they were to their task, whether relevant information could be retrieved and updated, and the usability and the aesthetics. This would hopefully inspire developers of hypertext systems and applications.
There were 4 hypertexts shown, ranging through entertainment to education. The first shown was Spaceship Warlock, a cartoon film-like application where the user can interact in fairly simple ways in a fantasy world. This had high quality graphics and sound, but was less "hypertexty" than the others - i.e. fewer choices at any point. The second was Cosmic Osmo, a collection of HyperCard stacks where the reader explores an unknown solar system. The hypertext is in black and white and the sound is low quality, but the humour and creativity with which the hypertext is put together more than make up for its simpler presentation. The third was much more an educational application about a Mozart string quartet (the Dissonant Quartet), where the student can explore the music of the quartet and the historical period, and even do a quiz. The fourth was Point of View, a system which students can use for creating their own hypertexts.
You will need to ask the participants whether our venture was a success or not.
(For the non-Americans, the title of the panel is a pun on the television ratings in the USA.)
Mark led a lively discussion, and was able to moderate as well as contributing to the discussion personally with the appropriate use of a cap.
Without trying to summarise the exact order of who said what, when, I state some of the points I found interesting or thought provoking.
Can hypertext disorientation be distinguished from bad writing? Producing linear material is also difficult but we are now used to it. A "linear" book is hard to navigate without the aid of a bookmark. Navigation needs to be designed into the material. Tools, such as tables of contents and indices, can be added to the material. Sometimes the user needs a map, or a guide, and sometimes the material itself is complex. Authors need to start with the assumption that readers will get lost, and so design for quick recovery.
There are different types of uses of hypertext, such as a novel, engineering documents, or students creating own material, and different amounts of training required, so you don't always need to design for a naive user. Each user has their own particular task.
Is retrieval fundamentally different from hypertext browsing?
Hypertext somehow promises you something, so when you feel lost you expect a solution. Are there ways of a system noticing that someone is lost - perhaps they visit the same node over and over.
Another way of looking at the navigation problem is to come up with designs for trying to disorient the reader, for example:changing the places links take you to; using bidirectional links without telling the reader why the link was created.
One interpretation of Pat Wright's talk (see p 4 of this report) was not that an extra click can be a significant cognitive load, but that people learn to use this new technology very quickly.
A search takes you to a seemingly random place in a hypertext, so Peter Brown created an animation of the route taken, but users didn't like the solution.
This panel struck a chord, since I have been sitting on the academic/industrial fence for a number of years now, trying to figure out the "real-world" problems which need to be tackled by the academic community.
The participants were a user, a vendor, a technologist and an academic (with apologies to all four participants for such brief descriptions).
There are no documented successful applications of hypertext, though it is difficult to define success. If time or money has to be saved then there are no examples.
There was slight conflict with commercial people saying that researchers don't look at "useful" problems, and academics answering that they would love to look at useful (read larger, more complex) problems if they were given sufficient money to do so. "Send money!" is how Frank Halasz put it.
Aircraft producers want to reduce the costs of their documentation. The paper by the Boeing group (see "Industrial Strength Hypermedia" on p 13 of this report) could change the research agenda.
One thought-provoking question (from Mark Bernstein) was :"If you buy a car and produce a compiler manual in 1991 which will be worth more in 3 years?" We have the notion that documentation produced now should still be valuable in three years, while we don't expect this of much "simpler" things like cars. Some documentation, for example for planes, does need to be accessible for many years, and it would be even better if we could reuse the relevant parts.
There is a fuller contribution in the proceedings.
People trying to cooperate in their work will want to act synchronously and asynchronously. How do these modes relate to each other, and how can you have a smooth transition between the two?
In one system writers had problems assimilating comments when they were linked to the original text. A later version displayed them next to the original text.
Planning tools are a good idea but they do disturb the content discussion (meta-level control).
The purpose of including summaries of each paper in this report is to give a brief overview of the subject matter covered by the proceedings. The length of each summary reflects my interest in the paper and not the quality of the paper itself.
A list of the technical requirements hypermedia tools must satisfy to be usable in an engineering environment is given.
The limitations of current products and requirements of future ones are:
· interoperability, over different platforms and with applications sharing a platform;
· shared workspaces, team members need to be able to work together;
· interactive authoring, multiple authors need to be able to interactively create and follow links;
· link and object attributes, users should be able to define their own attributes and values for links and objects within the environment;
· link anchors should be able to be created anywhere, regardless of data type;
· links should be able to have types and functionality associated with them;
· links need to be private, and public to different groups;
· templates should allow the automatic creation of links between regularly structured material;
· navigational aids, such as dynamic browsers, querying mechanisms and bookmarking facilities should be available;
· configuration control, mechanisms allowing users different access permissions for creating, deleting and modifying objects, links and webs are needed;
· concurrency control, authors should be given as much freedom as possible to modify objects even while others are reading them, or even modifying them;
· programmability, developers need a selection of tools for example to modify the hypermedia functionality or to manage hypermedia webs;
· operating systems, storage and networks, the technologies need to be developed to allow more efficient interprocess communication and remote storage of objects.
The paper explores the use of hypertext to enhance the process of component selection from software libraries, using tools in a prototype system called Kiosk.
The paper reviews the intellectual property scheme in Xanadu and contrasts it with current copyright law. Xanadu's predictions about reader and author behaviour are examined in light of how people currently behave in existing systems, such as computer conferencing and electronic mail. These analyses identify some respects in which intellectual property systems might have to be changed.
A hypertext index is enhanced by taking into account the context of the user and whether a particular item was a success or a failure for the user's current task. The model of the user's interests is progressively updated.
An aggregate is a set of distinct concepts that can be grouped together to form a more abstract concept. Two graph theoretical algorithms are used for constructing aggregates of hypertext nodes, thus creating a simpler structure for the hypertext. The hypertext links used are uni-directional. [The authors do not use typed links, and they have not checked whether their automatically created aggregates are comprehensible by users. Compare with "Don't Link Me in" on p 16 of this report, and "CYBERMAP", p 17.]
"In order to combine hypertext with database facilities, we show how to extract an effective storage structure from given instance relationships. The schema of the structure recognizes clusters and exceptions. Extracting high-level structures is useful for providing a high performance browsing environment as well as efficient physical database design, especially when handling large amounts of data."
Taxonomic reasoning deals with comparison and classification of highly similar nodes. This requires a broader view of hypermedia, not as nodes of information with links between them, but as a system of nodes through which users can move nonlinearly. The paper describes HyperSet, a set-based hypermedia system for supporting taxonomic reasoning.
[This can also be compared with the "Identifying Aggregates" paper on p 15 of this report, and "CYBERMAP" on p 17 .]
The authors look at different levels of structure in hypertext. They explore different methods for finding and representing, often hidden or implicit, structure on large, medium and small scales. They also look at tools for suggesting which places to explore next in the hypertext.
A way of placing windows in the context of an existing set of windows is presented. Hierarchically structured information can be displayed at various levels of detail.
The (textual) information in a plain or hyper text is analysed and a map of the structure is automatically created. Two classifications need to be carried out - one to determine which categories need to be created (called Hyperdrawers), and one to determine in which each node should be placed. A map is then created with the calculated structure of the Hyperdrawers shown - this has nothing to do with the original links in the information, but is based purely on the contents of each node. A demonstration has been produced in HyperCard. The author demands that the following requirements be satisfied:
· each node must have a well-defined place in the Cybermap;
· related nodes must be in the same Hyperdrawer;
· the distribution of nodes into Hyperdrawers should be well-balanced;
· the algorithm used to create the Cybermap must produce the same maps for the same input, and must be resistant to small changes in the boundary conditions and the user profile.
The current implementation produces only one Cybermap for the whole hyperdocument. A possible extension is to layer the Cybermaps, i.e. create a hierarchy of Cybermaps in Hyperdrawers.
The paper contains illustrations of the application. [Even though it is classified in the Presentation Issues section the paper could also have been in the Discovering Structures section. The author has not checked whether the automatically created Hyperdrawers are comprehensible by users. Compare with "Identifying Aggregates" paper on p 15 of this report and "Don't Link Me in" on p 16.]
A tool for flying through hypertext systems and the implementation issues are discussed. The tool, implemented in HyperCard, is similar to flicking through the pages of a book, but not necessarily in a linear order. This is an additional, not a replacement, navigational tool. The user can quickly travel through all the nodes in the hypertext using a number of strategies - e.g. depth first, breadth first, sequential. While flying, HyperCard messages are disabled, and so side-effects are avoided.
A prototype hyperspeech system is used to explore issues of navigation and architecture in an audio environment without a visual display. Speech recognition is used to manoeuvre in a database of speech segments; synthetic speech is used for control information and feedback. Recorded interviews were segmented by topic, and links were added to connect related ideas. Most of the boundaries between segments occurred at natural pauses between phrases, rather than between words within a phrase. All knowledge is embedded in the links and nodes. Two interfaces are discussed.
In voice systems, time, not screen real estate, is the most valuable commodity. The recorded voice segments are played back 25% faster, without a change of pitch.
Intermedia was extended to allow authors to define sets of pre-linked documents which can be duplicated - hypermedia templates. They describe some general principles for building consistent hypermedia collections, and details of the user interface to the template facilities.
Such a template facility allows the creation of consistent collections of documents. When copying template documents, both links between documents in the template, and links to external documents, are preserved.
The authors propose a construction kit which provides design objects, or building blocks, for three functionally distinct parts of a hyperdocument - content, organisation and presentation. The content nodes and links express the "natural" structure of the information. The organisation nodes and links allow different macrostructures to be imposed on top of the content network. Presentation styles can then be assigned to the different structures.
This model combines hypertext links with the concept of context nodes - a means of hierarchically organising documents. Different applications may observe the same node in different ways. The model offers operations to explore both structures of the hyperdocument - the links, and the nested context nodes.
Hypertext is viewed as a means of interfacing to dynamic information systems, such as decision support systems. A mapping from components in the information system to hypertext structures is introduced. The issue of document interchange of dynamic hypertext is also discussed.
"A technique is described for adapting the apparent structure of a hypertext to the behavior and preferences exhibited by its users while browsing." As a reader traverses hypertext links durations of events are altered, without changing the underlying links.
Intermedia was used in teaching the emergence of philosophy and science in the context of religious story and material culture. The experiences of the course organiser, the author, corroborate and extend the pedagogical impacts of hypermedia documented at Brown University.
The Chemistry Online Retrieval Experiment is studying the possibility of creating a useful, usable electronic library for chemistry researchers. Three systems (2 electronic, 1 paper) are compared for carrying out tasks representative of typical uses of scientific journal literature.
HDM, the Hypermedia Design Model, expresses information in terms of hierarchies of components , with different perspectives for each component and bodies for representing actual content. The basic HDM can be translated, manually or through a compiler, into a Dexter-type node and link model.
An object-oriented hypertext model using structured types to incorporate knowledge in hypertexts, is presented. Active behaviours may be associated with the types. The MacWeb system, which allows the creation of networks of information, is described.
Frames are used as a way of tightly integrating hypertext and structured knowledge representation. Partitions of hypertext nodes can be used for explicit representation of structure. The authors' experience suggests that hypertext provides a useful intermediary representation of knowledge between informal and formal.
These are informal sessions organised during the conference by participants to allow groups of people to get together to discuss specific issues in depth.
This was an information-distribution session rather than a discussion. However, it was useful to have the HyTime effort represented at the conference.
The HyTime standard "ISO/IEC DIS 10744" is currently "on probation". The last voting date is 10 April 1992. Comments need to be directed through the standards body in your own country.
The purpose of HyTime is to allow publishers a means of expressing hypermedia information so that it remains in a useful format for a long time. Newcomb's own interests stem from a desire to create courseware material that can be re-used on newer technologies. HyTime will not lock down how information needs to be structured.
HyTime is a prescription for writing DTD's (Document Type Definitions) for SGML (a meta data description language). It is a collection of BNF-like definitions. The creators have isolated the features of information which are useful to standardise, and not standardised the rest.
For further information the SGML SIGhyper group has many documents available for members. The easiest place for further information is probably the article in the November 1991 issue of CACM, vol 34 no 11, pp 67 - 83.
There were 21 demonstrations, of which I had enough energy left at the end of the day to see 6.
There are multiple windows, each window contains one page. A link goes from an anchor in a page to a whole page. Text windows are resizable. There are sound links, which are equivalent to synchronising the sound with a page (a sequence of pages can be presented). They will have Postscript layout so graphics can be included in text. A graphical anchor is created by overlaying an arbitrary polygon on the original picture.
LinkWorks is a set of navigation and linking services that can be used with DECwindows Motif. Links can be created between applications and used for navigation. A history of the user's trail is kept. The links are stored separately from the applciation data.
Microcosm is an open hypermedia system that supports the creation and following of generic links between applications. It is implemented as a collection of parallel communicating processes which can be split into two distinct types - document viewers and filters. The current version has been implemented in Microsoft Windows 3, with plans for future versions on Macintosh and UNIX platforms. (See the proceedings of ECHT '90 for further information.)
These are tutorials I did not attend, but purchased the notes for (the other notes seemed more dependent on attending the presentation).
The ACM Hypertext Compendium is now available for the Macintosh (HyperCard or ASCII), MS DOS (ASCII) or Sun workstations (KMS) from $100 - $495, depending on system and whether you are an ACM member. Contact Bernard Rous or Margaret Tuttle, ACM Publications Department, 11 West 42nd Street, New York NY 10036, USA, tel +1.212.869.7440 for more information.
World Wide Web is a means of using a single hypertext-like interface to access many existing data systems via existing protocols (FTP, NNTP) or via HTTP and a gateway. It was originally aimed at the High Energy Physics community. If you would like to try using this, or would like to make your own data available, then contact Tim Berners-Lee, Robert Cailliau, CERN,
CH-1211 Geneva 23, Switzerland, tel +41.22.767.3755, +41.22.767.5005, fax +41.22.767.7155
"Hypertext:The convergence of contemporary critical theory and technology", George P Landow, The Johns Hopkins University Press, Dec 91, 7 x 9, 240 pp, 10 illus, 0-8018-4280-8 $45 hardcover, 0-8018-4281-6 $15.95 paperback.
"From Memex to Hypertext:Vannevar Bush and the Mind's Machine", James M Nyce and Paul Kahn, Academic Press, Dec 91, 384 pp, 0-12-523270-5 $39.95.