Andreas Dieberger
Georgia Institute of Technology
Graphics, Visualization & Usability Center (GVU)
Atlanta, GA 30332
USA
andreas.dieberger@acm.org
© K. Andrews & A.Dieberger, 1996
Although a substantial corpus of hypermedia research was available to the designers of WWW in 1989 and 1990 and did indeed influence its design (Berners-Lee 1990) , many important findings are not reflected in the Web's current architecture. Practicalities necessitated trade-offs and compromises, the consequences of which are being felt today.
The main advantage of the WWW is its simplicity. Although standardization often restricts variability and sometimes flexibility the Web initially was "standardized" on such a simple level that many different applications of the Web are possible. However this simplicity also resulted in the trade-offs mentioned above. Recent additions to the WWW -- in part done to remedy these trade-offs -- endanger the simplicity and therefore the flexibility of the Web.
Keeping the Web as simple as possible definitely was one of the reasons why links are embedded in HTML and not kept separately. However as the Web grows we start to feel the limitations of this approach as it directly hampers efforts to create visualizations of Web structures, as it seldom supports structured authoring and as it does not permit us to do searches based on (non-existent) structures in the Web.
Closely related to this problem is the lack of meta-information on the Web. This also restricts the possibilities for adequate visualizations and intelligent navigation and search tools.
The Web has reached a size where it is a useful environment to work in because a lot of interesting information is out there. It is only difficult to find it because search engines may return not 5 relevant documents but rather 5000 mostly irrelevant ones. We are not even able to further process these search results because they are delivered in a non-standard format.
The two limitations mentioned above are a direct cause for such usability problems on the Web. That these issues still are not tackled adequately is a result of having lost the link between hypermedia research and the Web since the Web was originally designed. The lack of a strong link between the HCI field and the Web doesn't make this situation any better but rather leads to several other usability problems.
In section 2 we will focus on the two main limitations of the Web's design -- embedding of links inside documents and the lack of structuring facilities. In section 3 we discuss several other usability problems the Web faces today, namely the lack of user feedback in image maps, the lack of meta-information which (like the lack of structuring principles) restricts the usefulness of search mechanisms and visualizations and the fact that the availability of 'cool' features in Web browsers temps authors to make unreasonable use of it. We also mention the need for more natural ways to interact over the Web and to exchange pointers to information on the Web. In section 4 we touch issues of global user interface design for a World-Wide Web and in section 5 we summarize the paper.
From a user interface perspective, the chief (and often only) mode of navigation is the hypertext link. Most classic hypertext systems provide much more elaborate information structures which in turn support the navigation and the understanding of the hypertexts structure.
Even the earliest Web browsers provided navigational aids from the classical hypertext literature, such as:
Structuring facilities can be provided in two places: in the server (better) or in the authoring tool. As the Web does not provide structuring facilities itself we can only provide them in the authoring tools. Unfortunately as the Web does provide link typing only in later versions of HTML (and as most browser do not support it, nobody uses it) structures created in authoring system are generally lost after conversion to HTML.
The lack of typed links and structure makes it also impossible for search engines to determine boundaries of Web structures and, for example, to restrict a search to a certain hyper document. This is a typical example for how the Web can learn from traditional hypermedia research as most hypertext systems do provide structuring facilities.
External linking structures would allow people to navigate the Web based on its structure and not only based on the links defined by the author. Such an external linking structure would also be available for analysis, which then would allow to create visualizations of the structures, or structure based queries. Structure based queries were advocated as early as 1988 by Frank Halasz in his 7 issues for future hypertext system (see (Halasz 1988) .)
That the only structuring facility available on standard Web servers is the hyperlink inevitably leads to user disorientation (referred to in the hypertext literature as the "lost in hyperspace" syndrome) within the "World Wide Spaghetti Bowl" (to quote Robert Cailliau). Hierarchical structures can be simulated using menus of links, but maintaining these is difficult (and rapidly becomes a nightmare for hundreds and thousands of documents) and structural links are not readily distinguishable from standard associative links. For this reason, many information providers prefer to serve HTML documents from Gopher servers, thereby imposing a hierarchical structure on their information space (Lindner 1995) . An even better solution is to use a Hyper-G Web server, which combines hierarchical structuring with consistent hyperlinks and fully integrated search facilities.
The lack of structure on the Web is closely tied to the lack of standardized meta-information (like author information, modification dates, keywords and so forth). This is another other major stumbling block on the path to modern navigation tools and visualizations of the Web's structure.
Essentially this means that image maps are static images and the result of a click on them can be seen only after waiting for a result from the server. The first disadvantage of this method is the slow feedback users get for their actions. Depending on the server response time a significant amount of time may pass before any feedback is shown for the user's action.
It is also impossible for the client to show the anchors of links in such a clickable image as the contours of the anchors are kept on the server. Using client side maps at least the anchors can be shown, although the most commonly used Web clients don't bother to show this information. Another problem is that clickable images often are designed without showing any affordances, thus making invisible anchors an even bigger problem.
Ironically image maps are one of the few instances where the Web does make use of an external linking structure as advocated above: the links are indeed kept outside of the actual document. However this separation is realized in a way that it is impossible to provide immediate user feedback, which is a very important ingredient for a usable system.
James Seidman's client-side image map proposal (Seidman 1995) allows link anchors to be specified in the HTML wrapper around an inline image. These links also are not embedded in the picture -- they are defined externally to it but they are available to the client when the user interacts with the picture. Therefore they allow the client to show where anchors are located and to provide immediate user feedback. Although undoubtedly an improvement over ISMAPs, they still only apply to inline images and not to external documents or arbitrary document types.
Another way to provide user feedback is to use extensions to the Web like Macromedia's Shockwave, which allows authors to provide immediate visual and even acoustical feedback for user actions. Shockwave essentially plays Director movies embedded in Web pages (Macromedia ) . But such plug-ins go far beyond the actual Web and thus restrict the Web's simplicity. Also links are again embedded inside documents and cannot be edited any more once a movie is converted to -- for example -- Shockwave.
Yet another possibility it to provide feedback using Java script or Java (Flanagan 1996) . However his approach requires programming and it is very likely that we will end up with with many different, non-standardized solutions and documents or applets that contain hard-coded links embedded in them.
With standardized meta-information (like author, owner, title, size, creation date, modification date and so forth) it would also be possible to get much more useful information from search engines, that would allow to directly base visualizations on these results. Presently the only reasonable way to create visualizations for the Web is to use information retrieval approaches and to try to generate the missing meta-data out of the text of the Web pages and to extract the missing structure through an analysis of the Web pages' source text (see for example (Mukherjea, Foley et al. 1995) .)
Such an analysis is bound to be incomplete because many links are hidden in image maps or external files like Shockwave movies. Also pages that consist mainly of non-textual information often cannot be classified using such approaches (see next section.)
We all know the small notes on many Web pages like "Needs to be viewed with Netscape Vs. x and plug-in y installed" and so forth. While it is a good idea to provide a special experience for users who have these gadgets there should be an alternative for those who have not. Many authors don't bother to provide this alternative. Typical examples are beautiful graphical pages that are empty when looked at in a text-only Web browser.
One of the main problems in such a case is not only that there is no textual information available on these pages but also that the 'frozen' text in these graphics is not available to search engines. Should a tool try to create visualizations based on content it is very likely that such pages will be classified wrong as the relevant information is not available in textual form and as there is no meta-information to give the necessary information.
Especially with the current plethora of 'cool' features the Web also needs more intelligent servers than those used today. It is a bad idea to send the same picture or MPEG movie to all users -- even if they have no chance to adequately display the data. Not only would this reduce the used bandwidth, it would also avoid displaying huge pictures in a small browser window where the only chance to see the rest of them is by extensive use of scrollbars.
Another feature that may lead to unusable pages is free control of background and text colors (as if background GIFs were not bad enough.) Backgrounds are a useful tool when we want to provide visual site coherence -- a subtle change in the background can tell us that we leave a specific sub-area in the site and move to another. However many background / color combinations are violent attacks of our visual senses. It is not uncommon to see color combinations like a black background with bright green text. The anchors are dark blue (which is almost unreadable on the black background) and anchors that were traversed already are bright red. Not only is this color combination gruesome to watch -- it even communicates the wrong information because links that have been used already are much more prominent than unused links (the opposite of what is normal and intended).
Many authors seem to be oblivious to the fact that they probably cannot even read their Web pages themselves. They use what is 'cool' and not what is useful. This situation reminds us strongly of the early days of text processing where people -- suddenly able to use more than one font -- wildly mixed fonts and font sizes on one page. Obviously we did not learn from this. Provide users with cool features and they will use them to their own disadvantage. Introducing new features is good, but we need authoring tools that help users to use them in a reasonable way.
But there is more to this issue than a plea for a more reasonable naming scheme. There is a strong tendency to communicate URLs, be that in emails, advertisements, other media and so forth. The difficulties in handling URLs makes this communication a difficult undertaking. In order to really support easy communication of Web page locations we need tools that help us pointing out Web pages in a more natural way.
Actually one of the major shortcomings of the Web today is its lack of support to communicate about it. Although new tools for this are released almost every day they are still crude because they normally rely on manipulating URLs either by typing them over and over again or by copy/pasting them -- into emails, from emails, into forms, from HTML source text, from handwritten notes jotted down in a meeting or while watching a TV ad and so forth.
There are two ways to tackle this problem. One of them is to provide communication tools on the Web that allow to point out pages in a very natural way that essentially hides the URLs from the user (see for example the Juggler system (Dieberger 1996) .) The other solution is to create a more user-friendly naming scheme for Web pages, like symbolic names (as in Hyper-G) or URNs (Universal Resource Names) (Daniel and Allen 1995) .
The URL is a physical resource location rather than a resource identity, akin to specifying a book by its location on a particular shelf of a particular library rather than by its ISBN number. Aside from the problems communicating URLs this also leads to the known problems of dangling links and network bottlenecks. A working URN (Universal Resource Name) mechanism is therefore badly needed (Daniel and Allen 1995) for the WWW.
The current Web also is predominantly read-only -- servers provide and users browse. Better user and group access mechanisms are needed as well as on-line editing tools. This leads into the whole area of support for communication and collaborative work in the wider sense.
With the advent of Java (Java 1995) or similar technology the WWW then develops into a globalized user interface to diverse information providers and services. The name "World-Wide Web" hints already at the host of problems this will lead us into. Suddenly we do not only have to cope with the problems of providing a text that is understandable and acceptable for a global audience (the problem of icon design, the choice of color and appropriate backgrounds and pictures existed for the plain WWW already) but now we are talking global user interface design for applets and information pages. It is clear that a lot of work has to go into user interface issues to produce services that are globally usable. Hypermedia research traditionally had a strong connection with human factors research. This connection is still missing for the WWW.
Similarly many authors try using spatialized metaphors like drawings of offices or buildings, often in situations where these metaphors are utter nonsense and although it is commonly accepted that this type of representations often is a problem.
The amount of information on the Web is the reason why Web users spend less and less time on each page, frantically surfing around to find a page that captures their attention for more than a few seconds. Sites that are visually most intense are perceived as cool and valued most highly. This behavior often lures authors into using just the cool features of the Web and not paying attention to the design of their information structures or trying to create useful pages. People who use the Web mainly for work still seem to be in the minority (or they just are not as 'loud') -- a situation which does not enforce good design.
It almost seems as if the Web is bound to make all the old mistakes over and over again. The problem is that there is no place where Web authors could have learned it better. Everybody can make Web pages and only few people who teach page design stress the importance of information design and visual design (this is changing -- fortunately).
Therefore we need another approach to that problem. It will always be possible for HTML hackers to create incredibly ugly pages but in the near future most people won't want to bother with coding HTML by hand anyway. Instead they will use tools to create their Web pages. If we provide authors with tools that gently guide them towards reasonable design we are bound to see better designed pages on the Web after not too long. This may be a clear answers to the question of where traditional usability and hypermedia knowledge can be introduced to improve the Web.
That knowledge has to go into the authoring tools that should be more than just text editors. These tools should also help authors to create reasonable design not by restricting them but by gently guiding them into the right direction. They should also be tools that help people design information structures for the Web, complete with the necessary meta-information and typed links to show structure.
And there the another of the above mentioned problems of the Web ties in. How shall all this be done if there are no higher level structuring principles on the Web? This is very difficult. Therefore we need to provide higher order structures (like Hyper-G collections) on the Web that support users creating and organizing structures. This however would require to re-create the lost link between hypertext research and the Web.
Ayers, E. Z. and J. T. Stasko (1995). Using Graphic History in Browsing the World-Wide Web. Proc. of WWW4, Boston, accessible at http://www.w3.org/pub/Conferences/WWW4/Papers2/270/.
Berners-Lee, T. (1990). HyperText Design Issues, W3 Archive, accessible at http://www.w3.org/pub/WWW/DesignIssues/.
Bernstein, M., J. D. Bolter, et al. (1991). Architectures for Volatile Hypertext. Hypertext'91. San Antonio: 243-260.
Daniel, R. and T. Allen (1995). An SGML-based URC Service. IETF Internet Draft. accessible at http://www.acl.lanl.gov/URI/URCspec/.
Dieberger, A. (1996). Browsing the WWW by interacting with a textual virtual environment - A framework for experimenting with navigational metaphors. Proc. of Hypertext'96. Washington DC: 170-179.
Flanagan, D. (1996). Java in a Nutshell. Bonn, O'Reilly.
Haan, B., P. Kahn, et al. (1992). "IRIS Hypermedia Services." Communications of the ACM 35(1).
Halasz, F. G. (1988). "Reflections on Notecards: Seven Issues for the Next Generation of HyperMedia Systems." Communications of the ACM 31(7): 836-852. .
HyperG Welcome to Hyper-G, IICM. accessible at http://hyperg.iicm.tu-graz.ac.at/hyperg.
Java (1995). Java A White paper. accessible at http://java.sun.com/whitePaper/javawhitepaper_1.html.
Kappe, F. (1995). Maintaining Link Consistency in Distributed Hyperwebs. Proceedings of INET'95. Honolulu, Hawaii. accessible at http://info.isoc.org:80/HMP/PAPER/073/abst.html.
Lindner, P. (1995). Using Gopher with the World-Wide-Web. Proceedings of GopherCON'95. Minneapolis, Minnesota. accessible at gopher://boombox.micro.umn.edu:70/00/gopher/Gopher_Conference_95/Papers/WebbedGopher.
Macromedia Shockwave. see http://www.macromedia.com/Tools/Shockwave/.
Microcosm Microcosm. see http://cosm.ecs.soton.ac.uk/.
Mukherjea, S., J. D. Foley, et al. (1995). Visualizing Complex Hypermedia Networks through Multiple Hierarchical Views. CHI'95. Denver, CO, ACM Press: 331-337.
Netscape Netcom Home page. see http://home.netscape.com/real-index.html.
Nielsen, J. (1995). Multimedia and Hypertext - The Internet and Beyond. Cambridge, Academic Press.
Seidman, J. (1995). Client-Side Image Maps. Internet Draft. accessible at http://ds.internic.net/internet-drafts/draft-seidman-clientsideimagemap-02.txt.