In the late 60s and early 70s, we began to see real applications of computers to real information retrieval tasks. (By the early 1970s, computers were so commonplace that Green could write a paper discussing the indexing of specialist material, referring throughout to `data bases' or `data-bases', without once mentioning whether or not a computer was involved [71]!) At the same time, a number of themes emerged. Those that have continued in one form or another until the present day are dealt in later sections, but two that might not seem so familiar now are summarised here.

A major impetus to computerisation in information work lies in the tasks involved in printing and publication. Although there is some sense in which printed services are antithetical to computer-based services such as online, nevertheless the production of printed services was an early, fertile ground for computerisation. Thus Lynch [43] discussed a method for generating printed indexes, and later [57] reviewed a number of methods. The PRECIS system was designed with machine-generated indexes in mind (Austin [81]). Gralewska [54] described a system which included printed index generation; this system reflected what became a universal theme in operational systems, of using a database of machine-readable records, from one or more input processes, to generate a number of different forms of output (in this case both current awareness and retrospective search services, via printed output or machine searching). Similarly, Clough and Bramwell [67] described a package (ASSASSIN) intended for industrial use, which accepted input from tape services and allowed for printed indexes generation as well as machine searching.

Perhaps the most influential development of this period was that done by the major publishers of abstracts journals. Although this was not very well represented in the Journal (one exception is Hyslop's [39] discussion of the system from which Engineering Index was produced), there was an excellent review by Martin [76] of computer typesetting, which gave a very clear summary of why the production of such secondary services was so ripe for this development. Martin did, however, make one comment which, in retrospect, is remarkable for having been so wide of the mark:

... authors have suggested that the mere existence of large volumes of text in machine-readable form as a result of the increased use of tape-driven composing machines, with or without computers, could have a significant potential from the point of view of future information systems. In the present writer's view, this potential has been much exaggerated and is unlikely to be exploited...

In fact, it would be hard to exaggerate the effect that such availability did have. The market in the tapes of machine-readable data that were the by-product of this development was just starting in the early 70s (for example, the ASSASSIN system mentioned above was designed to take such input). Very shortly afterwards, two U.S. systems developed as in-house services started offering online search services to subscribers on a commercial basis, based entirely on such tape databases: Lockheed's Dialog and SDC's Orbit. Thus was the modern information industry born.

From the mid-60s until Ronald Reagan hijacked the acronym, SDI meant `selective dissemination of information', a phrase coined by H.P. Luhn. This was a form of current awareness service (i.e. helping the user to keep up-to-date with current publications), intended to be selective enough for an individual or small group, based on a `profile' of the interests of the individual or group. Much early work in computerised IR was directed at this end rather than at retrospective retrieval. One reason is probably that before the online era, retrospective searching by computer could be very slow (e.g. overnight or worse); current awareness is more obviously suitable for such hardware.

Corbett [47] ran SDI from a tape service; McCash and Carmichael [61] used an in-house database. Hall et al. [58] prepared a number of specialist current awareness bulletins. The ASSASSIN system referred to included an SDI function. Barker et al. [72] experimented with methods which might lead to automatic profile construction for SDI (an early form of relevance feedback -- see next section). Leggate [84] reviewed SDI services.

The idea of SDI enjoys periodic revivals of interest -- for example, there was a paper in the Journal in 1990 [132]. More usually now a different name is used -- recent examples have been `filtering' and `routing'.