Although the networking of medium sized computers has made headline news, and individuals and institutions have been urged to buy into a network as a way of achieving access to a number of computers, it is by no means clear that this will always be the best strategy. Quite recently a revolution has taken place in the mini-computer market. It is now possible to buy a moderately powerful computer for a relatively small outlay. Since information channels are likely to be routed through libraries for some time to come, it is interesting to think about the way in which the cheaper hardware may affect their future role. Libraries have been keen to provide users with access to large data-bases, stored and controlled some where else often

situated at a great distance, possibly even in another country. One option libraries have is the one I have just mentioned, that is, they could connect a console into a large network. An alternative, and more flexible approach, would be for them to have a mini-computer maintain-

ing access to a small, recently published chunk of the document collection. They would be able to change it periodically. The mini would be part of the network but the user would have the option of invoking the local or global system. The local system could then be tailored to local needs which would give it an important advantage. Such things as personal files, containing say user profiles could be maintained on the mini. In addition, if the local library's catalogue and subject index were available on-line, it would prove very useful in conjunction with the document retrieval system. A user could quickly check whether the library had copies of the documents retrieved as well as any related books.

Another hardware development likely to influence the development of IR systems is the marketing of cheap micro-processors. Because these cost so little now, many people have been thinking of designing 'intelligent' terminals to IR systems, that is, ones which are able to do some of the processing instead of leaving it all the main computer. One effect of this may well be that some of the so-called more expensive operations can now be carried out at the terminal, whereas previously they would have been prohibited.

As automation advances, much lip service is paid to the likely benefit to society. It is an unfortunate fact that so much modern technology is established before we can actually assess whether or not we want it. In the case of information retrieval systems, there is still time to predict and investigate their impact. If we think that IR systems will make an important contribution, weought to be clear about what it is we are going to provide and why itwill be an improvement on the conventional methods of retrievinginformation.