[TUHS] Unix NCP protocol stack for ARPANET

Paul Ruizendaal pnr at planet.nl
Tue Oct 26 18:06:30 AEST 2021


Noel wrote:

>> 2. Note that the BBN TCP worked over NCP as its primary transport.
> 
> Your terminology is confused. TCP _never_ ran 'on' NCP; they were
> _alternative_ protocol stacks on top of IHP (on the ARPANET). No
> AHHP, no NCP.

Yes, of course you are right. I meant BBN TCP used *Arpanet* as its primary transport and hence has drivers for the IMP interface hardware.


Lars wrote:

> Here's the rub.  Some hosts may have jumped the TCP/IP gun ahead of the
> 1/1 1983 flag day.  The host tables don't say.  Could it be that all
> those VAXen were running experimental TCP/IP in January 1982?

From Mike Muuss’ TCP digest mailing list and a mail conversation with Vint Cerf a few years ago I understood the following. “Flag day” wasn’t as black and white as we remember it now. During 1982 there was a continuous push to move systems to TCP, and over the year more and more systems became dual protocol capable and later even TCP only. Because all TCP traffic used the same, dedicated Arpanet link number, BBN’s network control team could monitor the level of usage. From memory, in the Summer of 1982 traffic was about 50% TCP and by October 70%. Presumably it reached 80-90% by the end of the year.

During 1982 on 3 occasions, network control activated a feature in the IMP’s that refused traffic on link #0, which NCP used to negotiate a connection. This caused NCP applications to stop working. Again from memory, the first outage was a few hours, the second one a day and third one, late in 1982, for two days. This highlighted systems that needed conversion and helped motivate higher ups to approve conversion resources. It seems that making the switch often involved upgrading PDP11 to VAX.

From what I can tell flag day went well, although there were issues with mail gateways that lasted for several weeks.

At the start of 1982 there was no (usable) VAX Unix TCP code that I am aware of. There were several options for the PDP11, but of those I think only the 3COM code worked well. Around March/April there was code from BBN (see TUHS 4.1BSD) and from CSRG (4.1a BSD). A special build of PDP11 2.8BSD with TCP arrived somewhat later. My impression is that this was still the state of play on flag day, with 4.1cBSD only arriving well into 1983.


More information about the TUHS mailing list