[ih] TCP adoption in 1984

[Lawrence Stewart]

Various additions enclosed.

Regarding the Xerox PARC and Systems Development Division (SDD) machines, the sequence was Alto I, Alto II, Dorado, Dolphin, and Dandelion.

The Alto was about 500 KIPS of BCPL instructions, IIRC.  The ECL Dorado was about 4 MIPS of BCPL or Mesa instructions, and sufficiently large and hot that they were kept in the computer room.  The Dolphin was originally intended as a very good printer controller, and as such was optimized for memory bandwidth rather than processing.  It was office-capable, but not as fast as Dorado and kind of too expensive for broad use.

The Dandelion was a low cost design originally by Lampson (I think?) and finished by Bob Garner and others at SDD. It turned into the 8010 Star machine.  

Regarding early Ethernet, Bob Garner has collected a lot of history about this, https://ethernetalliance.org/voices-of-ethernet/ might be a good place to start.

I think the microcode support for hardware in the Alto gave people a false impression of how easy it would be to build Ethernet controllers.  If I remember correctly, the Unibus Ethernet was a full hex card.  The first 3COM board was made to fit in a multibus form factor, but only by making it half duplex (I think).  Incidently, Xerox in El Segundo built a single-chip 3 MB Ethernet around 1982, although it would only run at half speed.  At PARC we used it for the Etherphone project, with an Alto gateway using a 3 Mb board with a half-speed crystal.  

For Packet Radio, the PRNet was mostly IP.  Xerox used it as a transit network for encapsulated PUPs.  This was a good test for fragmentation, but not very enlightening for internetworking. In fact I think the encapsulation was inside the PRNet Channel Access Protocol rather than inside IP - just to save a few bytes of the tiny MTS.

-Lsrry