[ih] Correct name for early TCP/IP working group?

[Jack Haverty]

Noel said: "Perhaps I have missed something, but it seems, in 
retrospect, that the only really significant change from TCP 2 to TCP 4 
was the TCP/IP split, and the creation of UDP." During that era, Vint 
moved the "Gateway" work from the Packet Radio group to the group I was 
managing at BBN. At the time, that included SATNET, which had previously 
evolved from a research project into an operational network. SATNET, and 
ARPANET, were both continuously managed by the NOC at BBN. TCP, and a 
collection of "core gateways" had become increasingly important to 
users, especially in the UK in Peter Kirstein's group, which had to rely 
on TCP through SATNET to access computers in the US. The charter for our 
new contract was to move the Internet "core" to operational status, 
i.e., "make the Internet a 24x7 reliable service", or something like 
that. The obvious way to do that was to adopt the NOC's techniques and 
tools that had been refined over a decade or so of operations. The 
Internet needed ways to diagnose and isolate problems, monitor activity, 
deploy new software, and use the network to perform such tasks even when 
it wasn't working well at all. Just like the ARPANET. To do that 
required adding a lot of technology to the base TCP system as defined in 
TCP2. Lots of stuff was "ported" from the ARPANET world, such as XNET. 
The concept of "fake hosts" from the IMPs was brought into the Internet, 
so that useful operations tools were available - things like traffic 
generators. Various IP "options" were defined, such as several varieties 
of "source routing", which were very useful when the normal routing 
mechanisms were failing, or when there was a need to test one of a 
particular gateway's interfaces that had been declared "down", by 
forcing traffic to use it. "Ping" was possible by implementing a simple 
"fake host" in a gateway or host. SNMP was defined, including facilities 
to manage not only IP traffic but also TCP, since that was where much of 
the error detection and response happens in response to failures that 
someone ought to notice and fix. In research environments, the equipment 
involved is usually located near the researchers, so such tools are not 
much of a concern to them. In operational environments, the equipment is 
often very far away and in multiple locations. Because of the pressure 
to get out a DoD Standard, much of that new "operations" technology 
didn't make it into the TCP or IP spec that Jon assembled. Rather, some 
of it was collected and named ICMP - Internet Control Message Protocol. 
Others just got their own documentation, sometimes in an RFC, sometimes 
just in email messages. That incompleteness of the specs caused some 
problems especially as DoD systems began to come online after 
traditional government contractors got involved. They had been tasked to 
implement the DoD standards. Their contract didn't say that they had to 
implement ICMP. So they didn't. That made it very difficult to actually 
operate a network based solely on the DoD Standards. We lobbied very 
hard to get the additional technology implemented in systems being 
deployed by the Army, Air Force, etc. I'm not sure what, if anything, 
changed in the "DoD Standards" regulations. One of our "operational" 
government clients tasked us to write a handbook on Network Operations. 
They couldn't find any such "User's Manual" in all of the RFCs et al. So 
I wrote a large document titled something like "How To Manage A Large 
Network". Sadly I didn't keep a copy but it may exist still, buried 
somewhere in DTIC or a government warehouse. So, while the split between 
TCP and IP was the most prominent change from 2 to 4, there were a lot 
of other changes to make the Internet "reliable 24x7", which continued 
well after the DoD Standard was created. There were a bunch of 
unfinished "research topics" that also weren't there yet - e.g., the 
desire for routing mechanisms based on actual transit time rather than 
hops. Jack Haverty
On 1/28/25 09:13, Noel Chiappa via Internet-history wrote:
>      > From: Jack Haverty
>
>      > My recollections, all IIRC of course
>
> Thanks very much for those; I had previously set out to improve the "TCP and
> Internet Meetings" page:
>
>    https://gunkies.org/wiki/TCP_and_Internet_Meetings
>
> a bit (I'd be very pleased to hear any comments about any remaining errors,
> and what else it needs), and I 'borrowed' a few chunks of your message, to
> explain the context, and who came - I hope that's OK.
>
> I too did live through most of this (the first meeting I came to was the
> August 1978 one). I just re-read most of the early minutes, looking for
> mentions of TCP 2.5 (didn't find much, alas), and most of the discussion
> seems to be about topics that later turned out to have been irrelevant, like
> EOL/Urgent and fragmentation.
>
> Perhaps I have missed something, but it seems, in retrospect, that the only
> really significant change from TCP 2 to TCP 4 was the TCP/IP split, and the
> creation of UDP.
>
>
> Note to future historians: there may be some detail errors in that message
> (well, it does say "IIRC"), so cross-check. E.g. "The IETF was formed to
> Engineer the operational Internet as it grew. The IRTF was formed to pursue
> the Research" - There was an 'InArc' formed at the same time as the IETF:
>
>    https://www.eecis.udel.edu/~mills/database/papers/inarc.pdf
>
> but it never went anywhere. I think the IRTF was created later - I'm not sure
> exactly when, perhaps around the time of RFC 2014? (I have this vague memory
> of the IETF and InArc being initially announced at the same West Coast
> meeting, around the time of IETF 1.) Maybe the IRTF was created to do what
> Inarc should have done?
>
> Another one:
>
>      > I had the impression that the INWG was part of the group that thought
>      > the datagram architecture was unworkable. Mentally, I associated it
>      > with X.25 and X.75 style of interconnecting networks. But perhaps that
>      > was a mistake.
>
> I wasn't there, but I get the impression that some considerable part of the
> INWG actually was sold on datagrams: Pouzin, whose CYCLADES/CIGALE was the
> key step from the ARPANET to internets, was a big player in INWG; and Cerf
> and Kahn's "A Protocol for Packet Network Interconnection" originally came
> out as an INWG document. See Alexander McKenzie's "INWG and the Conception of
> the Internet: An Eyewitness Account":
>
>    https://alexmckenzie.weebly.com/inwg-and-the-conception-of-the-internet-an-eyewitness-account.html
>
> for more detail.
>
> 	Noel

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20250128/3b41f71a/attachment.asc>