Introduction

Building new client software for interfaces to library systems is over. The user community will strongly resist another piece of software to view information. They want to view everything through the one interface, the browser. Attempts to modify this interface, such as the Java based Dynix client, while it may prove attractive features[BIBLIO] will be resisted by the casual user.

Libraries cannot control what equipment will be used to view their web sites. Their users will have a mix of equipment of different ages using different operating systems of different version. They will use different browser software with a variety of different versions and they may choose to turn off various features or override them. Libraries are not in control of how their information will appear on the users screen. With the stress on the remote delivery of service the lack of bandwidth to many patrons is an important consideration to avoid the "world wide wait" syndrome.

Normally a library will try and provide, as far as possible, equality of service to its various categories of patrons. This is not true of commercial sites.

Commercial versus library sites

We see many commercial sites with animated graphics which can only be viewed adequately with high end equipment. Why is this?

Commercial sites succeed by the number of people who view their site. For them to sell a product or service they need to attract people to the site and a pleasant 'viewing experience' is important. Also the people with money to spend tend to be those with the better equipment! Commercial sites are not there to reach everybody. They want to reach those people will generate revenue. This is especially true of sites they rely on advertising as their revenue source are paid on the basis of the number of visitors..

Libraries are not like that. They are close to natural monopoly for many of their services. They must succeed through content. Commercial sites succeed through design. This is not to say that content is unimportant to commercial sites and design unimportant to libraries only the stress is different.

Conflict between appearance and content

HTML is a markup language which, in its latest manifestation as XHTML[W3C], is moving to divorce structural markup from stylistic information, to separate content from display. Appearance is determined by the browser which may lack the capability to display what the designer intends. Also the user might deliberately disable some features. You cant make it look the same so why try? Yet web designer continue to spend vast effort on achieving the right appearance even using proprietary tags [SMITH] which only exist in a particular browser knowing it wont look the same anyway.

We are seeing the influence of designers who have transferred from a print environment and who believe appearance to be central. They approach content as a sculptor approaches a stone and feel that what they produce will be viewed by fellow artists when the viewers may be geologists. The print environment is one where the designer determines what is seen and there is no question that the viewer has anything but a static role to play.

But the designer does not exist in isolation. Approval processes for publications which require agreement from senior management before a publication goes out the door are still applied. But for print they are applied to a technology, print, which preserves mistakes whereas on the web errors can be modified as detected. The approval processes can be evolutionary and more relaxed.

Managers have another problem. Often they are older and have not the flexibility to cope with the new medium. They apply print habits of judgment to what they see. Their judgment might well be based on seeing a printed version of the page coupled with a demonstration on their local, and therefore fast, network. In the library world they may well be thinking of the flashier commercial sites they have seen, again on the local fast networks, and they wish to emulate them. What they should be doing is considering what the end user can , or might choose, to see. While I have no explicit evidence for this I suspect that:

The usability of a web page is inversely proportional to the degree if interest given to it by a senior manager.

Osterbauer and others at AusWeb2000 had some interesting comments to make on commercial sites. [OSTERBAUER]

We need a new paradigm in design where the author determines content and the designer (who may be the author) established guidance to enable the readers to provide a satisfactory appearance for their own needs. The end user should be enabled so as to easily modify the appearance of the document rather than being forced into the design decisions of the designer.

Rules of good web design

Nielson's 1996 paper is still a good start [NIELSON1996] and he has extended this [NIELSON1999]. His original list of what to avoid is:

1. Using Frames
2. Gratuitous Use of Bleeding-Edge Technology
3. Scrolling Text, Marquees, and Constantly Running Animations
4. Complex URLs
5. Orphan Pages
6. Long Scrolling Pages
7. Lack of Navigation Support
8. Non-Standard Link Colors
9. Outdated Information
10. Overly Long Download Times

In the context of libraries I would add to this list of what to avoid:

1. Graphics especially big graphics
2. Fixed pixel width tables
3. Tables embedded in tables for layout purposes
4. Reliance on colour to provide information
5. Scripting (or provide alternatives)
6. Plugins which are not supplied with browsers
7. Propriety tags

I would promote the use of:

1. Text equivalents to all images and multimedia via ALT tags
2. Headers for tables
3. Descriptions for tables
4. Summaries for graphs and charts
5. Standard HTML specifying the DTD used

A number of the items in this list are to ensure that the visually impaired can make use of the service and this should be a strong consideration of design for libraries. [W3C99]

Survey and methodology

The survey looked at the home page of each universities web site. Where the site used frames the page included in the main content page. The sites and URLs are in Table 1:

Reports

Reports were then run on these pages from:

The W3C Validation service [W3CV]

This service checks the html for validity and gives a detailed report on errors found.

The Bobby disability service [CAST}

This service checks to see if it is useable by the visually disabled.

The MacOS version of Tidy program [TIDY]

This program checks the html, corrects and reports on errors and improves efficiency style specifications. It provided a separate error report.

The reports were then run into fields of a FileMaker Pro data base and also the page source from the browser. The database can be obtained from the author should anybody wish to do further analysis.

Data extracted

The following data was extracted and added in separate fields of the database

• The size in bytes of each report and the source
• The number of errors encountered from the Tidy report and from W3C
• The download time for the whole page (this includes the graphics) at 28k baud in kbs.
• The total size of the page and associated graphics

From this a number of additional measure were calculated which are discussed in the analysis

The lack of accurate, or any DTD statements, created difficulties. In some cases fewer errors might have been reported had the DTD been accurately specified.. It would appear that some sites which started at a lower level of HTML, when developing the site, introduced HTML tags from later versions. These would be reported as formal errors.

Results and analysis

The numeric results are presented in Table 2.

Use of graphics

By and large graphics in a home page are for decorative or design purposes. Even in the case where they are used for navigation text alternatives would be just as functional. Generally a large proportion of the information to be downloaded is in graphical form averaging 71% of the content. Griffith University with 0), University of Tasmania and La Trobe managed to keep below 25%. University of New England, Australian Catholic University and Royal Melbourne Institute of Technology managed to go over 90%.

Download times

The download time on a 28k link averaged at 20.3 seconds with a standard deviation of 10.7. Griffith University, University of Tasmania, Monash University, La Trobe University, Flinders University and University of Wollongong managed to keep under 10 seconds. In the spirit of World Wide Wait Queensland University of Technology, Macquarie University, Edith Cowan University, Royal Melbourne Institute of Technology and Australian Defence Force Academy managed to get over 30.

Quality of HTML

Thirty one out of the 37 libraries were reported by Tidy as appearing to use HTML proprietary tags rather than conforming to the HTML standard. This is clearly unfortunate.

Only 13 actually used an SGML DOCTYPE header to show which HTML version they were using although some of these specifications were invalid.

An error measure, the average number of errors from Bobby and Tidy per kilobyte of source code, gave an average error rate of 4.3 errors/K with a standard deviation of 6.8, a wide variability.

University of Melbourne, University of Wollongong, University of Sydney, Australian National University, University of South Australia, Royal Melbourne Institute of Technology, University of Adelaide, Macquarie University, Charles Sturt University, Queensland University of Technology and Griffith University managed less than one error/K;

Australian Catholic University, Curtin University of Technology, University of Newcastle, Central Queensland University and University of Tasmania were over 10.

Frames

Only two sites used frames. One, ANU, solely for a graphical effect which could only be discerned by a careful eye.

Scripting

Scripting is often a source of problems with old browsers and with variant versions of scripting languages or proprietary extensions such as those by Microsoft. Scripting should be used with care. Eighteen sites used scripting. W3C reported errors in 16 of these sites in relation to the scripting.

Meta tagging

One would hope that libraries would pay attention the meta tagging and 31 did but sadly only one used the Dublin Core standard.

Disability testing

None of the libraries achieved clear "Bobby approved status" . But 18 libraries only required manual checks of potential problems which the bobby program could not assess automatically. The other 19 libraries clearly failed the test.

Correlation?

There seems to be little correlation between the variables examined but more detailed work is needed. A spreadsheet was used to normalise the data for each variable. The normalisation expressed the data as the number of standard deviations from the mean value for the data set to a value of zero. Plotting normalised error rates against the normalised graphics percentage gave the result in figure 1.

Figure 1

A figure of merit

Taking average of the normalised scores for the error rate, graphics percentage and download times we can use this as a somewhat arbitrary figure of merit

The top 10 institutions on this basis are:

Griffith University, La Trobe University, Swinburne University of Technology, Flinders University, James Cook University, University of Queensland, Charles Sturt University, Monash University, University of Wollongong and Northern Territory University.

The bottom 10 on the other hand are:

Bond University, University of Western Australia, Macquarie University, Australian Catholic University, Curtin University of Technology, Edith Cowan University, Australian Defence Force Academy, Royal Melbourne Institute of Technology, University of Newcastle and Central Queensland University

Conclusion

Australian university library web sites, based on their top page, are disappointing and they all fail to varying degrees. While individually they do not have all these failings the following statements would apply to many of them:

• They place an over reliance on graphics (Griffith being a notable exception)
• Many of the download times are too long to be comfortably used over a phone line.
• Their HTML used does not conform to HTML standards and contains a significant number of errors
• They fail the standard usability tests for use by the visually disabled

Sadly most of these errors could easily be avoided

References

[BIBLIO]

Biblio Tech Review: Information Technology for Libraries. Dynix WebPAC http://www.biblio-tech.com/html/dynix_webpac.html

[CAST]

Centre for Applied Special Technology. Welcome to Bobby 3.2. http://www.cast.org/bobby/

[SMITH]

Josh Smith. Browser Compatibility Chart Oct. 29, 1999 http://webreview.com/pub/1999/10/29/feature/index3b.html

[NIELSON1996]

Jakob Nielsen. Jakob Nielsen's Alertbox for May 1996: Top Ten Mistakes in Web Design. http://www.useit.com/alertbox/9605.html

[NIELSON1999]

Jakob Nielsen's Alertbox, October 3, 1999: Ten Good Deeds in Web Design http://www.useit.com/alertbox/991003.html

[OSTERBAUER]

Christian Osterbauer et al. Web usability testing: A case study of usability testing of chosen sites (banks, daily newspapers, insurances) http://ausweb.scu.edu.au/aw2k/papers/osterbauer/paper.html

[TIDY]

Clean up your Web pages with HTML TIDY. http://www.w3.org/People/Raggett/tidy/

[W3C]

W3C User Interface Domain. HyperText Markup Language Home Page . http://www.w3.org/MarkUp/

[W3C99]

Web Content Accessibility Guidelines 1.0 http://www.w3.org/TR/WAI-WEBCONTENT/

[W3CV]

The W3C Validation service. http://validator.w3.org/

Copyright
http://conferences.alia.org.au/alia2000/proceedings/tony.barry.html