• Steve Summit maintains a FAQ for the C language, which he's also published in a book version. His page has the information.

• There is an eclectic collection of material and pointers to material about C and other languages at the Lysator Society site hosted at Linköping University in Sweden.

• If you're interested in old Unix systems beyond the papers saved here (in particular, trying to get them or run them, whether on on real hardware, via simulation, or just for study) look at The Unix Heritage Society site, and the nearby PDP-11 Unix Preservation page.

  The current nominal proprietor of the source code for early research Unix systems (5th Edition through 32V) is named SCO. Today's SCO was previously called Caldera, Caldera having bought the rights from the earlier instance of SCO. This earlier SCO made them available for research and personal use. Until recently, Caldera did as well; now their offer has been withdrawn, presumably for reasons having to do with their suit against IBM. An earlier (2000) version of their license has evidently been withdrawn; this is a local copy. There was also a subsequent, less restrictive, license; its announcement is shown here in PDF format.

  In the past, Caldera/SCO coordinated with the TUHS and PUPS group. The actual sources are well-mirrored in various places.

  The PUPS site has various emulators, so these systems can be run without a physical machine.

• Tom Van Vleck maintains a fine archive of material at www.multicians.org about the Multics system.

Some worthwhile sites that annotate and collect even older material are listed here. They often cross-reference each other, but are big. These are some that I've found worth exploring.

• Karl Kleine maintains an excellent collection of documents (language manuals, early patents, machine descriptions): things like the Algol 60 and A68 reports, various Modulas, and on and on.

• Manchester University's celebration of their history. I would particularly recommend looking at the article in Nature magazine of 22 October 1949, which anticipates, almost eerily, themes seen in computer architecture today. To excerpt:

  An important feature of any storage system is its accessibility ratio

  alpha = t1/t2,

  where t1 is the time taken to read a chosen address of the store, and t2 (>= t1) is the time elapsing between the demand for that information being made, and the end of the reading process.

  Clearly, it is necessary that the part of the storage to which access is frequently made should have a value of alpha approaching the optimum value of unity, if the time taken in solving a problem is not to be extended by barren periods of waiting. In the electronic storage system¹ used, alpha = 1. However, it is unnecessary to have alpha = 1 for all the storage, and, in fact, a magnetic storage system, for which alpha = 1/133, is used to supplement the electronic system.

  The "electronic" storage system was a Williams CRT storage tube; the "magnetic storage" system referred to a drum memory, which (in this paper) is later footnoted with a reference to unpublished work by West and Thomas. We have here both the notion of a storage hierarchy, and, one suspects, some hardware that wasn't quite working yet.

• The Virtual History of Computing at Oxford; it contains many fascinating links.