[ih] Did the MIL-STD matter? [IETF relevance (was Memories of Flag Day?)]

[Brian E Carpenter]

On 05-Sep-23 05:22, Jack Haverty via Internet-history wrote:
...

> Separately, there were efforts, initiated by someone, to orchestrate the
> "Flag Day" on the Arpanet, to declare TCP/IP a DoD Standard, to define
> and implement a formal certification program for new TCP
> implementations, and probably other efforts I never knew about.
> 
> Someone was in charge, and someone was doing lots of things to "make it
> work".
> 
> It wasn't perfect.   Actually it was a bit chaotic IIRC.
> 
> For example... Jon Postel took on the task of documenting TCP/IPV4 so it
> could be referenced as a Standard.  RFCs were released.   DoD declared
> them mandatory for all military systems that involved communicating
> computers.
> 
> A bit later, at BBN we were assisting various pieces of the government
> in getting their computer systems up and running with their vendor's
> brand-new, certified, standard TCPIPV4s.  It was a big surprise to
> discover that, although TCP/IP was there, none of the other "tools" we
> had been using for years had been implemented on those machines.
> 
> Much of that missing functionality was called "ICMP", well documented in
> RFC 792.  But only TCP/IP had been declared a DoD Standard.   

This was more important than you might think. The first TCP/IP document
I ever saw was MIL-STD-1777, and I think that was not uncommon in Europe.
At the time when a major argument for OSI was "because it's an ISO
standard, all the vendors will implement it", the fact the TCP/IP wasn't
just some academic research thingie but was a mandatory US military
standard was a very powerful argument. It made TCP/IP respectable at
management level, not just something that a few Unix geeks were keen on.

Certainly when I took over the CERN networking group in early 1985,
although we expected OSI/CLNP to rule the universe, MIL-STD-1777
was physically lying on my desk. When I had to walk back CERN's OSI
policy in the late 1980s, "TCP/IP is a US MIL-STD" was still a
useful argument.

> Government
> contractors, who had not been involved in the research community, had to
> implement the Standard.   But the Standard didn't include ICMP.  So they
> didn't implement it.
> 
> That made it much more difficult to "make it work".  For example,
> without ICMP as the Internet's Swiss Army Knife, you couldn't even
> "ping" a DoD Standard computer.   I remember we raised quite a fuss
> about that, and implementations started to appear.  I'm not sure if the
> Standard was ever modified to require ICMP.
> 
> Other things, like SNMP, were useful but also missing.  Many people
> apparently didn't consider ICMP and its cohorts to be part of TCP/IP.
> We considered such technology essential to be able to "make it work".

Nevertheless, TCP/IP "made it" and OSI, which had its own version of the
same problem, didn't.

      Brian