The SIGIR project Saving the Old IR Literature has scanned and released a new batch of historic IR (Information Retrieval) papers, including early papers on the SMART system and papers on the development of test collections. The papers are written by amongst others: Gerard Salton, Karen Sparck Jones, William Cooper, Keith van Rijsbergen, Stepen Robertson, Martin Kay, Michael Lesk, and Nicolas Belkin. The new batch is listed below and available from the SIGIR web site.
The collection contains some unique documents, for instance Karen Sparck Jones' and Keith van Rijsbergen's Report on the Need for and Provision for an 'IDEAL' Information Retrieval Test Collection written in 1975, which I anxiously searched for when doing my Ph.D. research. The document is an important mile stone towards the current TREC conferences; work that already started in 1960 with Cyril Cleverdon's Cranfield experiments, one of Computer Science's earliest examples of empirical testing in a laboratory setting.
It's all there, enjoy!