On Thu, Mar 15, 2018 at 5:04 PM, Dave Horsfall <dave@horsfall.org> wrote:
On Thu, 15 Mar 2018, Tony Finch wrote:

As I understand it, the old hosts.txt registrations got grandfathered into the DNS in the .arpa zone - they were ARPANET hosts (e.g. see RFC 921). The modern structure was set up after the transition to IP, so it's fair to call .com and friends Internet domains. (See RFC 920.)

Yeah, I was referring to DNS as opposed to HOSTS.TXT of course.

But it looks like there were a bunch of .edu and .gov names before Symbolics.

I'm happy to be corrected/updated; I happen to have an interest in geeky history (as if that wasn't apparent).

​Well this history is sort of strange because it was more random/back patched than the historian likes to admit.  For instance DEC definitely had an ARPAnet connection and I think IBM did also. Note Tektronix and HP did not have ARPAnet connections, but Tek was a very early UUCP site and HP followed suite about 2 years later.

As others have pointed out the ARPANETname space was a tad more flat:  user@site and SITE was sort of large (MIT, CMU, DEC) - which originally mapped to IMPs.    Different networks (like SAT-NET) started to strain the naming scheme. As Ron reminded us, with the coming of the splitting off the DoD to DDN's responsibility and the creation of MIL-Net people began to talk about needing more that 'site.'

Hence the creation of responsibility 'domains' ​-- which (as I recall) was less for SW naming and more for administrative control.

Their a a number of salient points here.  First there was no real registration as we think of today.   For instance, I was personally assigned ccc.com because I had been 'ccc' on the UUCP net, and I was already moving packets around from UUCP and to Arpa/Internet folks via gateways.

It was confusing time and a lot of different networks had bridges/email gateways.  At that time, my friends at BBN just put me in the database long before I was directly connected or assigned my own Class C network (which happened a few years later).   This allowed ARPAnet, CS-Net etc to send emails to me the gateway point was defined in the BBN database for MMDF (not sendmail BTW) -- I've forgotten where all that name washing got done -- it might have been BBN proper, but somehow I think UDEL was in the middle of some of that [someone like Ron might remember]. 

Adding sites like me was done because BBN was trying to 'flatten' some of the UUCP naming mess on their side (the message on the 'user' part of the ARPA addresses was a wash with funny characters - like !, % etc to try to slime in the different gateways].

A key thing to remember here, is that all happened before Symbolics was incorporated.  And I was 'late' compared to others.   In fact 'Tektronix' was running IP internally but used UUCP for external email as they could not get an ARPAnet connection.  They later would joined CS-NET and used Phone Net when that was first allowed and finally they were an early 'commercial' IP site that BBN connected with Cisco gear.

It should also be pointed out that since DEC was existed on the ARPAnet and was grandfathered as DEC.Com (and I believe IBM the same) - this is all long before Symbolics.Com was created.  So it's hard to really call Symbolics 'first' other than to say the were probably the first firm to ask BBN how do we do this and make it formal, right around the time when DoD was trying to get out of the ARPAnet and was trying to find a way to transition it to commercial firms.

For whatever it is worth, I also remember being ticked off when I got my first bill years later from Network Solutions because at some point, the BBN/SRI databases came under their control.   I had been CCC.COM for a long number of years by then (5-10 maybe) and what was this all about.