[ih] TCP adoption in 1984

[Bill Nowicki]

Thanks for Noel, Larry, etc. to fill in other details of this era. Just one more Dorado anecdote, I promise. As a summer intern at PARC in 1981, I worked on some early packet voice experiments. Larry Stewart mentioned EtherPhone, of which this was the experimental test. Detractors said that Ethernet was too unreliable to be used for real-time communication. Xerox had me write some simple code (using hardware that I think Larry designed?) to interface an Alto to an analog telephone and we had fine conversations over Ethernet. Then another bit of code to record and play back the audio. The Alto did not quite have the power to do the networking and store it on disk at the same time, so I used a Dorado for the recording. 
At the time, someone estimated that given the number of top-level computer scientists and engineers who worked on the Dorado, the small number produced, the enclosure affectionately called the "armored personnel carrier", the cost of running dedicated keyboard-video-mouse lines to each office, the expenses of power consumption and cooling, it worked out to close to $100,000 each.  I told my parent my summer job was to make a $10,000 computer act like a $20 telephone, and a $100,000 computer act like a $100 tape recorder.
Back to the original 1984 question, even by then Sun Microsystems was making about a thousand machines per year, doubling every year. At least about a quarter of those were used internally, om the TCP/IP network that was only connect via email relay to the wider Internet. Commercial customers (more than half the business) built similar "Intranets". SGI was doing the same thing a year later, as well as the other wanna-bes. Probably soon after if not in 1984, due especially to the PCs also on such networks, my guess was that the number of these "Intranet" hosts easily outnumbered the Internet routable ones. But I am curious if anyone did an in-depth analysis of this.
Bill