Origin: I typed in most of GNUBIB, but I had help from some other people, who allowed me to incorporate their personal bibliographies, or helped with the typing. GNUBIB also contains a lot of stuff that we got from the net; none of this latter stuff was copyrighted. The initial impetus was that one guy here had a good bibliography on logic programming (my field) published as a book, and he restricted on-line access to it because of the publisher's copyright. John and I set out to duplicate it; we ended with more than twenty times the number of references he had, because we did not restrict it to just one field. A short description of the collection: GNUBIB has over 55,000 references, distributed across just about all areas of computer science. Some areas are more represented than others: this reflects John's and my areas of interest. The well-represented areas are logic and functional programming, programming languages in general, operating systems, software engineering, computer architecture, parallelism, computer science education, and others I can't remember at the moment. Underrepresented is numerical analysis, real-world applications (whether of the General Motors or the Internal Revenue Service type) and some areas of AI such as expert systems, vision and robotics. GNUBIB also contains a little non-computer science material, e.g. the contents of Scientific American. GNUBIB is organized accourding to source: e.g. all journal references are in one directory and all references to a particular journal are in one file. There are a couple of exceptions: some journals are split across multiple files (they would be too big to edit otherwise), and there is a miscellaneous directory. About half of the references are to journals, and about half of the rest are from proceedings. For some journals, e.g. CACM and SPE, and for some proceedings, e.g. POPL and ISCA, our collection is complete (i.e. we have references for all the papers ever published in CACM since volume 1 in 1958). Zoltan.