[ih] Date of RFC 791 for celebration

[Bob Braden]

  *> 
  *> This, alas, predates me slightly, and I have the history somewhatl muddled in
  *> my head. 

Noel,

Don't worry about it.  I was there, and my memory is somewhat muddled.

  *> 
  *> And then there's the issue of version numbers: there are 2, 2.5, 3, 3.1 and
  *> 4. What I hear is that 2.5 was a "implementation split", in which the unified
  *> single header remained, but the code was split into two.

That is correct.  Curiously, you are the first person besides me who
remembers it this way (I remember it vividly, splitting my IBM OS/360
TCP/IP code.)

 3 was the first
  *> version that had the headers fully split (and included "protocol numbers" to
  *> identify which transport protocol was being used - see list below), but my
  *> guess is that it included variable-length addresses.  I seem to recall that

My memory is that no defined version contained variable length addresses.
Jon Postel and Danny Cohen, especially, argued forcibly for VL addresses,
and the rest of us might have gone along, but Vint over-ruled this.  His
argument was that to make TCP/IP acceptable to the DoD (which meant,
insulate TCP/IP from being swallowed by OSI) it had to be simple and
straightforward to implement in hosts and routers, and to him that
meant FL addresses.

  *> 3.1 had the variable-length addresses removed, and 4 was an editorial cleanup
  *> of 3.1
  *> 
  *> So:
  *> 
  *> - Is that the right order for the split, and variable-length address
  *> 	removal?
  *> - Can we identify the version numbers (2.5, 3, 3.1, etc) which go with
  *> 	each version? 
  *> - Can anyone identify the INWG meeting at which the latter happened?
  *> 
  *> 

I can perhaps do so, when I have time.  But as research progresses,
things are not always so clear cut.

Bob Braden