********** TECHNOLOGICAL CHALLENGES OF THIS DEVELOPMENT ********** ========== TOOLS FOR STATIC MATERIAL ========== Conventional CBT often quoted as 100 hours per hour of material, reduced with good tools. The cost of creating good Web material can be as great. Need ubiquitous tools - editors(list), imagemap creators, validity checkers (Pitcrows work, MOMspider), etc. Or converters of existing material (latex2html, frame2html, ms word macros, etc). ========== SELF PACED TESTS - READILY AUTO-MARKED ========== ---------- COMPATIBILITY, POSTABILITY - NEED FOR A STANDARD ---------- To share each others work, avoid duplication of effort, minimise start up cost for later colleagues. Similar to need for CGI ---------- AUTHORING AND MAINTAINABILITY ISSUES ---------- All in one file, parsed by the server, or in separate files? (possibly pre-parsed (examples of these approaches)) ---------- STATISTICS GATHERING AND INTERPRETATION ---------- ---------- MAINTAINING STATE ---------- This is not necessarily the same issue as adding state to the http protocol or adding a stateful method to the http protocol. The protocol defined the communication between client and server; I am here talking about the system of client, connection and server as a whole. The client is already keeping state, if you like - a document is sent out which contains links, and the client deviously keeps this information to allow t hose links to be selected! To say nothing of hotlists - oh, there is state being retained without a doubt! Joking aside, some form of state clearly needs to be maintained to allow marks for different users to be tracked. ---------- INTERACTIVE RESPONSE ---------- To some extent this can be maintained by faster networks, faster browsers, and lower latency methods such as MGET. The burden of assuring adequate response however still rests on the author. In some cases, it may be preferable to send a script, such as a macromedia director script, and have a player program on the client side deal with timeliness and synchronisation issues. The abstracted results - the score, for example, or other data about student performance deviously gathered - can then be sent back toi the server, perhaps by POST or perhaps by another method entirely - FTP? Mail? ========== SELF PACED TESTS - HARD TO AUTO-MARK ========== All the above issues, but an even bigger authoring one. How to specify the correct answers in some succinct format? How to standardise this? Realistically the markers are all going to be developed in AI labs around t he world; they will get their authors papers, but will not be compatible with one another :-) ---------------------------------------------------------------------------------------------------------------------------------- [[Chris Lilley]]