2.1. Traditional approaches
Much CBL material is either not formally distributed or, where it is distributed, floppy disks are still the most common media used. There is however growing recognition of the opportunities for electronic distribution of material that is already in an electronic form.
Using a network to distribute programmes has traditionally depended on file based service in particular ftp (file transfer protocol) has been the usual approach within the Internet community. For an ftp transfer the user needs to know the file name and the site name where the file is being stored. Then a series of commands must be given to transfer the file to the client machine before the installation process takes place. This itself can be made difficult by the need to reduce the number of files stored on the server machine and to hold files in compressed format.
Some of the problems of traditional ftp storage have been addressed by providing central indexes of archive machines (archie) and by designing client programs that allow much easier browsing and downloading of files (for example Fetch, xgetftp). There is also growing research into how the many large stores of files can be searched and whether search agents can be improved .
2.2. CBL collections
For the provision of learning material as distinct from other files there is another approach that should be considered. This is the building of collections in a way to encourage the use of the material contained on them. As an example material should be described, include examples of how it might be used and a guide to the installation process. Then access can be given to files for downloading and installing. While much of this is often offered in README files on ftp servers, the newer forms of server allow this to be presented in an attractive and seamless way. Recognition must also be given to any licencing conditions associated with the software and control held over those permitted to have access.
2.3. Data transfer
CBL material is multimedia in nature, that is it will consist of executables produced in various ways and for various platforms. Within those executables different media types, in particular graphics and increasingly sound and video sequences are likely to have been used. This means that files for CBL will often be fairly large and of different application types.
The centralising of files on servers can lead to bottlenecks in the transfer of data over the networks. This can be limited by local strategies such as mirroring of data to closer sites and caching of data when it is fetched from remote sites. Such automated strategies are not without problems. An example cited in  discusses an Australia based server which shows a large range of files on other sites, when these files are accessed they are transferred from the original site, cached at the server and passed on to the requesting client. However the wide range of files apparently available has led to the use of this server by clients all over the world often leading to unnecessary transfer into and out of the Australian network. This example emphasises that caching strategies need to consider how the system will be used.
An alternative to the transfer of programs over the network is to enable the control of local programs from the centralised description. Only if the program was not installed or had changed from that already placed on the client machine need it be transferred from the server.
2.4. Client and server communication
Control of client programs has been considered within the TLTP funded Interact project . The method adopted in that project is to adapt simulations written for the client machines to respond to signals from network browsing programs (in practice Mosaic ) and transfer data from the remote server, via Mosaic, to the simulation. In this way examples or previous data can be shared. This is being extended to allow the state of the simulations to be stored at any point and passed across the network. The Interact Communication Facility (ICF), suggested in this system, also avoids the security problems that would occur if arbitrary shell commands were passed between the server and the client .
A similar approach, again based on WWW servers and clients, is being investigated within the TLTP BioNet project . There a launching protocol is defined and a Windows based client, W3launch, produced to respond to remote commands from the server. The use of an intermediary program allows the user of the client machine to ensure that only approved programs are run, a command can only be carried out on the local machine if it has been entered into the list of permitted commands. Adoption of an approach similar to this but extended to include the installation process should limit the need for programmes to be repeatedly transferred from information servers.
These last two examples recognise the common case where installation of programs happens rarely and is under the control of an administrator. This is an important case but increasingly students are encouraged to explore the opportunities available over networks for themselves. If remote courses are to be made widely available in this way consideration should also be given to less limited modes of access.
2.5. Interaction and feedback
Within Computer Based Learning it is often argued that interaction should be encouraged. This applies also to the serving of CBL. Providing information on a large scale also allows the gathering of feedback data from a large number of users. Information servers now have the technology to allow interactions with users and this would typically be in the form of questionnaires to receive comments. The opportunities should also be explored more fully so that software may be customised more closely to the needs of the user based on an initial question and answer session. At a simple level this might be to ensure that the user is provided with the most appropriate form of the program for their platform. Others possibilities include that the program could be altered to operate at the most suitable level of difficulty and with the data files most relevant to the user's subject area.
The management of any feedback and questions asked about the programs can be made more efficient if the information gathered is ordered to best advantage. The Answer Garden concept  considers the use of a database containing previous questions and answers and the association of experts with each part of the structure. A register of experts will be very attractive to a growing server of learning material. This would allow managed access between those developing and implementing the CBL material and those using the material.
2.6. Summary of needs for CBL serving
The requirements for CBL serving can be summarised as