CHI 96 Workshop: HCI and the Web, Position Papers
Jan Erik Moström
Computer Science Department, University of Luleå
jem@sm.luth.se
Summary
A group at the Centre for Distance-spanning Technology at the University of
Luleå, is currently involved in publishing a regional newspaper on the World
Wide Web (the Epaper project). We are emphasizing the "usefulness" of the
paper and trying to create a newspaper that is easy to read, simple to
navigate and contains valuable extra services. As the platform we chose to
use "standard" HTML, but already in the initial phase of the project we
found several shortcomings of HTML that make it unsuitable for this type of
application. For example:
- Limited page layout control
- Difficulty creating an easy-to-navigate structure
- Limited interaction
- Slow and variable response times
- Low utilization of the computer's potential
Because of these shortcomings we have not been able to produce a newspaper
that meets our expectations. One possible tool for solving these problems is
the Java programming language. We plan to use this language in the next
phase of our project.
When we remove the hype currently surrounding Java we can see that is just
an ordinary programming language with some special properties that make it
particulary suitable for use on the Internet. Since it is "only" a
programming language, there is also the problem with badly designed user
interfaces.
Since Java is a new language and it is not commonly used there are few
guidelines and examples of how good HTML/Java pages should be designed. We
see a clear danger of finding many badly designed pages in the near future.
HCI people can help avoid these problems by:
- Creating good examples of how HTML/Java pages should be composed.
- Giving guidelines for how distributed systems should work from the
user's perspective.
- Helping to create tools for building HTML/Java pages, tools that can be
used by the non-programmer.
Most of the problems are not new or unique, but can be found in the
literature. There are also new angles of old problems that need to be
studied closer. We think that the HCI community can make a valuable
contribution to the development of the Web using "old" knowledge and some
new research.
Introduction
The general goals of the Epaper project are to:
- Produce an electronic version of the Norrbottens-Kuriren newspaper.
- Investigate and try different methods of presentation.
- See what kind of "add-on" services are suitable.
During February 1996, we have produced a test version of Norrbottens-Kuriren
that uses "standard" HTML only. During the development of the test version
we have encountered several problems both with using computers as media in
general and HTML as the representation of content.
Why HTML is not Enough?
Page Layout
HTML is by definition a "structure describing language", and therefore, pays
very little attention to the layout of information on the screen. At the
other end of the spectrum we have languages like PostScript that are "page
description langauges", they emphasize how the layout will look when
displayed but pay no attention to content.
Using HTML in an heterogenous environment like Internet is good idea since
it allows the browser to display the content in the best possible way
available on a specific hardware/software platform. However, when producing
a publication like a newspaper layout is an important part of the
"signature" of the newspaper. In other words, the editor of the newspaper
wants precise and extensive control of the layout. Systems such as Adobe
Acrobat[TM] can fullfill this desire, but they normally introduce new
problems such as long download time and limited interaction.
What is needed is a way of controlling layout and presentation while
retaining the possibility of displaying information in different ways
depending on the platform.
Navigation
As with any hypertext system there is always the risk of the reader getting
disoriented (the "lost in hyperspace" phenomena). We already have this
problem in a small degree in our simple test paper. We can see that the
problem will increase in later versions as we add capabilities of following
a news story in time, searching, etc.
Our problem is not only to create a easy-to-navigate structure for one
issue, but we also have to support navigation in time and other different
strategies that will arise later in the project.
Limited Interaction
Standard HTML only allows for limited interaction. It is possible to enter
some information into a form and send that to a server who then sends back
new information. Even if this can be considered to be interaction, it is
very limited.
We would like be able to use techniques like direct manipulation in our
newspaper. For example:
The government has passed a new tax law that affects you in some way. In a
paper based newspaper you would get an table with some examples of how the
law would affect different people depending on a family situation, where you
live, income, financial status, etc. In a HTML based newspaper you could get
a table, a diagram, or the possibility to fill in a form. Then, you send it
to a server and get an individualized reply.
However, we would like to give the readers a diagram where they could change
the different parameters by dragging curves, icons, etc., and get an
immediate response.
There are other examples of how increased interaction would increase the
"value" of an electronic newspaper.
Slow and Variable Response Times
We have also noted that the variable and sometimes very slow response time
is another source of problems. This is especially noticeable when the user
follows a link that points to another server, perhaps in some other part of
the world.
We can also to see how readers sometimes spend several minutes reading an
article then follow a link to another article. They have to wait for several
minutes while the computer downloads pictures and text. It should be
possible to make a guess of what the user will do next and pre-fetch the
information. This is not possible with standard HTML.
Low Utilization of the Computer's Potential
It is also worth noting that HTML utilizes a very small part of the
potential of modern computer. It seems strange to use powerful computer and
be satisfied with publications that do not offer something new. If computer
based publications are to succeed they must offer something new and not only
mimic what can be found in books, videos or today's simple presentation
programs.
How Java Fits
One way of giving the publisher greater freedom to solve the problems
described above is the use of the Java programming language. Java makes it
possible to include programs on ordinary HTML pages. The programs are
executed at the readers computer and make it possible to do any kind of
presentation the publisher wants.
Why HCI People are Needed
Java is just another programming language similar to C++, and thus, it
requires substantial programming expertise to use. It also allows the
programmer to do really bad user interfaces. This is a big danger today
since there are very few good examples of how HTML/Java pages should be
designed.
Our experience is that the people doing Java development see that there is a
problem, but they do not know how to solve it. HCI people can give valuable
help to these people by using experiences from other similar applications
and with some new research.
Another important part is the complexity of programming Java. HTML and the
Web have been very, very successful in part because it is simple to create
some kind of Web page that can be viewed by a large audience. For many
people Java would be a step backwards since it requires so much of the
author/programmer. To make it short, the Web needs a good authoring system
that can be used by non-programmers. In creating such a tool the HCI
community would be an valuable asset.
CHI 96 Workshop: HCI and the Web, Position Papers