[ih] Internet-history Digest, Vol 2, Issue 16
clemc at ccc.com
Wed Nov 6 07:32:21 PST 2019
Let me see if I can rephrase a little bit to help add clarity to the
question. In my mind:
- To be '*The Internet*' it must be using a protocol based on the IETF's
IP family for interconnection between the hosts.
- To be '*an internet*' packets need to flow (in both directions)
between different hosts via some sort of connecting point/network splice.
Thus to you use your examples, Xerox's PUP style network and the original
UUCP based network really were internets***.
But an open question in my mind crops up when you switch protocols at the
splice point (*e.g.* the role ucbvax played between UUCP and the Internet's
IP based hosts). At the splice point, history shows that there was some
amount of each network 'leaking' into the naming/specifications/notations
of the other side (for instance, all the cruft sendmail had it for header
transformation and host naming).
By the previous definition these still meets the second point of packets
following through a networking splice point. The question is since
notions of each style of network bleed into the other side, at one point
have the two networks really become a single interconnected network? I
think that really becomes the crux of the argument and sadly the
difficulties presented by the bleeding of the notion of one into another,
is likely to have a bit of personal taint to it.
*** FWIW: Not to put too fine a point on it, Just I hate seeing people miss
punctuating 'Internet', I also find it distasteful seeing people calling it
'USENET' -- as the UUCP network predated net.news as the BTL internal
network - 'USENET' does not come about until BNEWS and CNEWS started
their rien as applications on top of it, but that's a different story.
PS I mei cuplea -- dylexia-is-me. I said Noah in my earlier message and
On Wed, Nov 6, 2019 at 9:59 AM Vint Cerf <vint at google.com> wrote:
> i think a lot of the interconnections were application layer gateways -
> such as email relays.
> Early USENET wasn't Internet if by this we mean TCP/IP based. UUCP was't
> using TCP/IP
> either. Xerox argues that it had multiple LANs running PUP and that is a
> kind of internet,
> but not using TCP/IP.
> I think I am still inclined to argue for the PRNET/ARPANET connection as
> the earlier of the TCP/IP style internets but even then the initial
> implementation was TCP only (with gateways re-routing based on the Internet
> addresses embedded in the TCP packet format)
> On Wed, Nov 6, 2019 at 9:52 AM Clem Cole <clemc at ccc.com> wrote:
>> On Tue, Nov 5, 2019 at 7:00 PM Dave Crocker <dhc at dcrocker.net> wrote:
>>> On 11/5/2019 2:38 PM, Joly MacFie wrote:
>>> > When was the first actual inter-network message sent using packet
>>> > technology?
>>> Might be interesting to seek some agreement about the biggest milestones
>>> for creating what we experience as the Internet.
>>> First, what are the criteria for a milestone? Conceptualization?
>>> Demonstration? A degree of production operation? Mass market adoption?
>>> Second, what are the easy milestones: packet switching and TCP/IP are
>>> obvious. What others? (I'm entirely biases towards wanting major
>>> applications to be added but, well, I'm biased.)
>>> Third, what are some less obvious but still essential milestones? I'll
>>> suggest NSFNet because it enabled both a standard for multiple
>>> backbonbes and an operational approach to infrastructure that became the
>>> foundation for the commercial Internet.
>> A very good question, i.e. I think you nailed it. You need to agree on
>> what an '*internet*' is before you can start to define when '*The
>> Internet*' was birthed.
>> Frankly, I'm not sure what the right answer is here as it was an
>> evolution and I'm not sure if there was any one particular event (like a
>> dinosaur kill off from an asteroid strike) that we can enumerate. But I
>> think I can postulate some other things that might be defined as an
>> 'internet' and I suspect other on this list can offer other examples, too.
>> e.g. As a minimum using TELNET, SUPDUP or the like, I know that CMU and
>> MIT built something internally to connect local hosts and allow them to
>> connect to the directly connected ARPANet hosts that had IMP connections.
>> CMU called this work the "distributed front-end", which replaced the
>> original "front-end" that was the directly connected glass tty's attached
>> to ASYLs on a PDP-11 which was also connected to each local ARPA host with
>> a DR-11B (the problem with the original FE implementation was each PDP-11
>> was connected to specific set of hosts so if you wanted to talk to host,
>> you needed a line to that specific FE). IIRC MIT used ChaosNet protocols.
>> For the DFE we used LSI-11 and original built something that was
>> ethernet-like (which we called ethernet at time but was a local hack) but
>> eventually morphed to 3M Xerox when we got access to the Xerox board and
>> transceivers (but started out as a local hack). FWIW: The CMU
>> distributed front end was originally implemented on LSI-11, but around
>> 1976 switched to Multibus 8085's and later after I left Stanford SUN
>> boards ??I'm guessing 1980/81??.
>> Someone from MIT like Noah can explain more, but IIRC: the Chaos stuff
>> ran on UNIX, LISP machines and much wider set of HW.
>> FTP and email was not allowed in the first versions, just remote
>> terminal, so we can argue that it will not complete inter-networking
>> solution. I personally think an important aspect of more formal 'internet'
>> is that things like the original DFE was basically unidirectional and the
>> hosts on the 'CMU side' were not exposed. So I think that somehow that
>> idea of packets flowing both ways needs to be in the definition of a full
>> That said, once an early IP stack started to appear, the distributed
>> front-end started to look more like a modern router as more and more
>> support for it went into each 'host" - it became a very similar in
>> architecture to the original CISCO AGS -- *i.e.* each local LAN became
>> its own network. Once that was done, support for things that exposed the
>> remote host to the other network became possible and things like email/ftp
>> As I said, MIT did the same sorts of things with Chaos and I think went
>> farther than we did early on to be honest. I also I think Stanford had
>> something similar, but I never knew the folks involved.
>> Also for the sake of argument, what about the UUCP network? It did not
>> support remote login, but it did support remote file transfer, email and
>> remote execution of jobs? It was bi-directional, al biet since routing was
>> explicit, was much harder to use than the ARPAnet protocols. My question
>> is, when hosts like ucbvax, decvax, or some of the other later CSNet hosts
>> 'bridged' - does that count as an internetwork?
>> Internet-history mailing list
>> Internet-history at elists.isoc.org
> New postal address:
> 1875 Explorer Street, 10th Floor
> Reston, VA 20190
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
Internet-history mailing list
Internet-history at elists.isoc.org
More information about the Internet-history