The saying goes that you can write a FORTRAN program in any language. Similarly, there are many documents on the web that are either very old, or look as if they are. Lacking any hypertext structure beyond perhaps a Previous and Next link, these linear documents are among the poorest uses of the Web. Of course the content may be excellent; it is the educational technology that is under discussion here.
Idle browsing of such uninspiring material is naturally ineffective in an educational context. There is no opportunity for interaction, the student is a passive consumer.
Forced structure such as Gopher gives the illusion of freedom but the outcomes are predetermined, the student still controls nothing.
Documents on the web have the potential to improve on a book. Pros and cons of linear material presented online - access from your desk, somewhat searchable (current file only) but tiring to read, why bother. Example eg IOP material??
Online materials should offer something you can't readily get from paper:
Subject guides, collected examples to encourage exploration. Example of the vis2 resource list.
Influence of comprehensive searchable indices such as WWW Worm. Saved searches as a form of current awareness on the cheap.
Active data - graph it yourself (mathematica links, application data types, MIME, the chemistry mime types. Online program source, the briefly existing Fortran 90 form at NAG)
The advent of imagemaps and forms has allowed the student to communicate with the server, and hence the presenter. This opens the way for self paced assesment integrated into the educational web (by using web or alternatively by using a hypermedia interactive application as a viewer and posting back the results)
Server log analysis to find mean reading times for documents, whether graphics were turned off, which external images were downloaded. Correlation with scores obtained in tests. This gives trainer feedback to help improve the course.
Feedback on each question as it is completed. Simulated example.
Adjustment of difficulty path through a question set based on running average score to stretch brighter students and avoid discouraging less able ones.
Live interaction with real people, by video conferencing, MOO and other means can be a valuable adjunct to educational webbing provided the number of students is small and they all live in similar time zones. Example GNA.
Text entry and evaluation. Equation entering. Drawback of doing this badly - the cheap adventure game scenario. Needs good AI and/or a restricted language and problem domain.
Points schemes and so on to encourage participation (examples GNA, Confer) but well designed and presented materials often motivate students without resorting to these tricks. (Is this a temporary effect due to the newnress of the web? Will students become habituated?)