Languages of the Net
Peter H. Salus · ACM SIGPLAN Notices · 1998
Forty years ago, there were no programming languages: then, toward the end of 1957, there was FOR-TRAN.A decade later, most people used that or COBOL or ALGOL or Lisp or APL.One of the important things to note is that all of these were procedural languages.They all consisted of functions operating on data.Pascal, Basic and C are further examples of such procedural languages.When the ARPANET was being conceptualized and then designed, one of the reasons for ARPA's investment was to link a number of different computers, each of which had a different proprietary operating system running on different hardware.At a meeting in Ann Arbor, Michigan, in late 1967, Wes Clark, then of Washington University in St. Louis, suggested that the future network connect self-standing mini-computers and that each main machine connect to its local mini.Thus, each site had to work out just how to get bits to and from the mini, not how to translate bits to and from other "hosts."These minicomputers were known as IMPs -Interface Message Processors.A contract to describe the IMP network was given to Elmer Shapiro at the Stanford Research Institute; the IMP contract was given to BBN in Cambridge, MA.They decided to build the IMP around a Honeywell 516.In December 1968, Elmer produced a report which was classified by the US Army and only distributed in numbered copies.This study of the design of a computer network made a number of assumptions, beginning with this first map of the ARPANET.Among other things, Shapiro discussed "The resources that a host has to offer to other hosts."These were:• Its collection of programs • The data in its files * Its capacity to do computing; and• Its capacity to store information.Elmer noted: "Clearly only programs and files can be transported among hosts," Remember: this was 1968.