[ih] TCP adoption in 1984

[Karl Auerbach]

Your question is perhaps more complex.

In 1984 MIT (John Romkey, Dave Bridgham - students of Saltzer and Clark) 
published PC/IP, a TCP stack for PC-DOS machines using the 3COM 3C501 
Ethernet card.

(We also saw, perhaps slightly later in time, TCP for the Macintosh by 
the folks who went on to create Intercon, and for the PC Beame and 
Whiteside in Canada, WRQ also in Canada [who I think were the first to 
push the Winsock API].)

PC/IP was a simple TCP intended mostly for use with telnet sessions with 
a human typing - which is why that implementation ignored window values 
sent by the remote host (the assumption being that there would always be 
an adequate receive window for data flowing at human typing speeds.)  
And that 3C501 was one dog of an ethernet card - a single buffer that 
couldn't be turned around from send to receive very quickly, so it often 
missed responses from quick hosts.

https://web.mit.edu/saltzer/www/publications/pcip-1986.pdf

PC/IP was rather popular.  The 3COM ethernet cards were a limitation, so 
PC/IP was updated to use Romkey's Packet Driver API to ethernet cards,  
  I think I wrote the first packet driver - for the ULANA procurement 
for the TRW Ethernet card that we created to meet ULANA's COTS 
[Commercial Off The Shelf] requirement].  Russ Nelson (Crynr) produced a 
mountain of packet driver implementations for a wide variety of Ethernet 
cards, most notably the rather popular Novell NE2000 cards.

PC/IP was turned into a rather more solid TCP suite, PC/TCP, by FTP 
Software.

All of this suggests that there may have been a lot more TCP going on 
among university folks with PCs and Macs, and among Silicon Valley folk 
than would be accounted for by a census of officially "attached" hosts.

It was during this period that we learned that if one had an Ethernet 
with PC's that one should avoid putting a Sun workstation on the same 
coax - we always suspected that the Suns were playing a bit loose (or 
more perhaps properly "fast and loose") with the Ethernet 
collision-sense/back-off timings so that they would win contention 
fights with machines, mostly PCs.  Add a busy Sun to a working Ethernet 
with PC's and the PC's would effectively go offline.

Life on Ethernet got so much better when Synoptics came out with twisted 
pair ethernet (which later evolved into the standard 10-base-T.  
Synoptics, and even more, David Systems saved the Interop net in the 
1988-1990 period more than once.)

         --karl--

On 5/2/26 3:26 PM, Bob Purvy via Internet-history wrote:
> Grok's estimate is that about 900 hosts could be reached via TCP, at the
> end of 1984. RFC 984 for TCP over Ethernet was in April of that year.
>
> Were these *entirely* university and research machines? Were any of them
> actually running TCP on Ethernet? Does anyone know"
>
> Thanks,
>
> Bob