Path: utzoo!attcan!uunet!lll-winken!xanth!ames!apple!bbn!bbn.com!mckenzie
From: mckenzie@bbn.com (Alex McKenzie)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Impact of BSD on the Internet
Message-ID: <40132@bbn.COM>
Date: 18 May 89 15:33:05 GMT
References: <8905162047.AA04059@dsys.icst.nbs.gov>
Sender: news@bbn.COM
Reply-To: mckenzie@labs-n.bbn.com (Alex McKenzie)
Organization: Bolt Beranek and Newman Inc., Cambridge MA
Lines: 65



Dear Root Boy Jim:

I'm writing in response to your message to the tcp-ip list titled
"Impact of BSD on the Internet".

1.  There was indeed a thing called NCP around in the early days of the
    ARPANET.  NCP was the acronym for Network Control Protocol, and it
    was the official host-to-host protocol from 1970 through 1982 (the
    official cutover data from NCP to TCP was January 1, 1983).  NCP had
    nothing to do with BBN;  it was developed by a committee of network
    host organizations called the Network Working Group (NWG), the first
    chairman of which was Stephen Crocker of UCLA.  I believe the first
    mention of NCP in the public literature was a paper by Stephen Carr,
    Stephen Crocker, and Vinton Cerf at the 1970 Spring Joint Computer
    Conference titled "Host-Host Communication Protocol in the ARPA
    Network."

2.  The first TCP/IP implementation for UNIX was written by BBN with
    Defense Communications Agency (DCA) funding.  It was written for
    UNIX Version 6 by Michael Wingfield and was completed by March 15,
    1979 (see IEN 93*).  Another early TCP implementation for UNIX
    Version 6 was written by Digital Technology Incorporated at about
    the same time.

    BBN also wrote the first TCP for Berkeley UNIX, with DARPA funding.
    The project was led by Rob Gurwitz and is described in IEN 168
    (January 1981) as being "designed for the VAX, running VM/UNIX, the
    modified version of UNIX 32/V developed at the University of
    California, Berkeley." As might be expected in a TCP project carried
    out by BBN, performance was optimized for the characteristics of
    wide-area networks.  The folks who were both starting SUN
    Microsystems and also directing the development of Berkeley UNIX
    wanted a TCP implementation optimized for LANs, and were successful
    in having the BBN TCP implementation removed from BSD releases and
    replaced with their own implementation which was so optimized.

3.  TENEX was a paged, virtual memory, time sharing system for the DEC
    PDP-10 computer which was developed by BBN to support AI research.
    It used paging hardware designed and built by BBN.  The development
    was funded by ARPA and TENEX became operational in early 1969.  It
    is described in a paper by Bobrow, Burchfiel, Murphy, and Tomlinson
    in Communications of the ACM, Volume 15, Number 3 (March 1972).
    TENEX served as the basis for DEC's TOPS-20 operating system.

    DEC PDP-10s generally, and TENEX specifically, were by far the
    single most popular "server" computers during the first several
    years of the ARPANET.  (In those days ARPANET hosts were commonly
    characterized as "users" or "servers"; a server provided file
    storage and cycles, while a user system primarily provided access to
    the network.) For example, in December 1970, according to a quick
    count I just made, 10 of the 20 ARPANET servers were PDP-10s and in
    June 1975, 23 of 47 servers were PDP-10s.  Although I do not have a
    breakdown of operating systems in use on these PDP-10s, I believe
    that over 3/4 of the DEC-10s on the ARPANET at any time were running
    TENEX.  Other popular operating systems for the PDP-10 were DEC's
    10/50 system and MIT's Incompatible Timesharing System (ITS).

I hope this information is helpful.
Alex McKenzie

* IENs are available online from the ARPANET Network Information Center
  at SRI-NIC.ARPA