[COFF] ARPAnet now 4 nodes

[gingell at computer.org (Rob Gingell)]

On 12/4/2019 8:19 PM, Larry McVoy wrote:
 > I'd love to know the order of nodes joining and how that was scored.

There are a number of sites that contain fragments of the history of 
sites/nodes joining the ARPAnet. Wikipedia's entry for ARPAnet has some 
of that along with several logical maps of the network (though these 
stopped including host #'s pretty early on.) One decent concise 
accounting is at https://www.zakon.org/robert/internet/timeline/

A collection of maps of the ARPAnet over time is available from the 
Computer History Museum at 
https://www.computerhistory.org/collections/catalog/102646704

As others have noted the numbers were assigned chronologically starting 
at UCLA's Sigma-7 system (1) such that the first 4 were:

1: UCLA: Sigma-7
2: SRI: SDS-940
3: UCSB: IBM 360/75
4: Utah: PDP-10

I know Case-10 was node 13 (more colloquially referenced as 15, in 
octal, which was how we saw the numbers and updated the host table). 
I've forgotten at this point whether the assignments were documented in 
RFCs or other assigned numbers documents from the Network Information 
Center (NIC).

 > I've been told that UW-Madison "was the 11th IMP on the arpanet" but I'm
 > pretty sure that is not true.

You're correct that it's not true. Not sure when/if UW-Madison was on 
the ARPAnet. Hosts did get decommissioned and I think the numbers were 
eventually recycled so it's possible a later site got a lower number but 
now it's my turn not to be sure.

 > My guess is that there are the original
 > IMPs that were arpanet, then there was an expansion to educational sites
 > and Madison was 11 on that.  Or something like that.

ARPAnet "vs." educational sites isn't a distinction that existed. The 
first 4 nodes consisted of 3 educational institutions and a 
university-affiliated research organization (SRI). In 1971 when there 
were 23 hosts and 15 IMPs more than half of the organizations were 
universities though contractors and labs started appearing pretty 
rapidly. IMPs apart from ARPAnet weren't a thing really.

There's a podcast, "50 Things that Made the Modern Economy" that had a 
fairly recent episode about IMPs that was pretty interesting. Had a 
economic, Adam Smith-like spin on the emergence of IMPs as an example of 
specialization in an ecosystem which was a big part of "why IMPs": the 
network was kept homogeneous among the IMPs and whatever weirdness was 
associated with the different hosts (mostly) contained to the host-IMP 
interface.

Those first 4 nodes were all pretty different systems. Some differences 
percolated up to the applications protocols, TELNET's options set was 
complicated by the heterogeneity of the hosts in a protocol architecture 
pre-dating things like "presentation layers" in network models.

If you peruse the maps at the Computer History museum site you can see 
some real diversity in the systems. For instance in April 1971 Burroughs 
had an IMP and a B6500 front-ending the under-construction ILLIAC-IV 
(later moved to Ames and front-ended by a couple of PDP-10s). PARC's 
MAXC appears in the mid-1970s. One of the goals in creating the ARPAnet 
was to provide access to unique resources to a wider research community 
and you can certainly see a lot of unique systems in those early maps.