[ih] Date of RFC 791 for celebration

[David P. Reed]

The TCP/IP header split was agreed to at a late fall or early winter 
meeting in Marina del Rey when I was still the MIT rep to that working 
group (perhaps that is the Jan 30-31, 1978 meeting, but my memory places 
it in 1977 for a good reason - it was not that close to the "blizzard of 
'78") .  Clark took over my role in this group sometime in the spring of 
1978, as I recall - which is when I was bearing down to complete my 
Ph.D. thesis and interviewing for jobs, so I had to disengage.

There was a significant difference between the prototype implementations 
that Tomlinson and others were testing (prelim code) and the specs that 
were being revised in the meetings on whiteboards, etc.   I wouldn't be 
surprised if there was a lag of months, but not years between agreements 
reached in meetings and docs relating to coded tests.

The meetings included reports from implementors on implementation 
issues, but also explorations of issues that went beyond 
implementation.   Jon's meeting notes would be helpful.

I somehow doubt that any one "birthday" makes sense to capture a process 
that was (in typical Internet style) an evolutionary, experimental, and 
creative process.
Lots of loose ends were being sorted out.

Fragmentation, for example, kept floating in and out of discussions and 
specs as the IP layer formed up after being separated from TCP (which 
had fragmentation from the start).   But the issue was *very* real and 
considered very important at the time, because nets with small packets 
were expected to be common (e.g. X.25 was enjoying huge hype).  (that 
did not turn out to be as big an issue as people thought).


Noel Chiappa wrote:
>     > From: Jack Haverty <jack at 3kitty.org>
>
>     > Have the "TCP Meeting Notes" that Jon Postel distributed after each
>     > quarterly meeting survived in electronic form?  These documented in
>     > pretty good detail the gyrations of TCP 2.5, 3, and 4 over the 1977-1978
>     > period.
>
> Would these be the same as these IEN's, I assume?
>
>  64  Sunshine   12-Mar-78   TCP Meeting Notes - 12 March 1977
>  65  Postel      5-Aug-78   TCP Meeting Notes - 14 & 15 July 1977
>  66  Postel     21-Oct-77   TCP Meeting Notes - 13 & 14 October 1977
>  67  Postel      8-Feb-78   TCP Meeting Notes - 30 & 31 January 1978
>  68  Postel     27-Jun-78   TCP Meeting Notes - 15 & 16 June 1978
>  69  Postel      9-Oct-78   TCP Meeting Notes - 18 & 19 September 1978
>  70  Postel     15-Dec-78   TCP Meeting Notes - 4 December 1978
>
> As far as I know, none are available on the web...
>
>     > E.G., there was a "Proposal for TCP 3" by Ray Tomlinson distributed at
>     > the 21 Oct 77 meeting of the TCP Working Group.
>
> That doesn't appear in the IEN Index, and a web search doesn't turn it up either.
>
>
> However, searching for it turned up something valuable this site:
>
>     http://www.cs.utexas.edu/users/fool/cs370/
>
> and in particular a link to this site (which I knew of, actually, but had
> forgotten about):
>
>     http://www.cs.utexas.edu/users/chris/think/digital_archive.html
>
> which contains a number of otherwise-unobtainable IEN's.
>
> So, for instance, it has a copy of IEN-21, "TCP 3 Specification" from January,
> '78, which allows us to see that TCP/IP 3 had:
>
> - Separate IP and TCP headers
> - Variable length IP addresses
>
> - There are smaller differeces (e.g IP - no checksum, no fragmentation
>   support; TCP - 32 bit ports, support for segments), but in general the rest
>   of it looks moderately similar to TCP/IP-4
>
> So it looks as if my prior take:
>
>     >> 3 was the first version that had the headers fully split (and included
>     >> "protocol numbers" to identify which transport protocol was being used
>     >> ..) but my guess is that it included variable-length addresses.  I seem
>     >> to recall that 3.1 had the variable-length addresses removed, and 4 was
>     >> an editorial cleanup of 3.1
>
> is likely accurate (although I wish we could recover a 3.1 spec to be sure).
> (Note: the field called "Format" in the IP3 header is what we now call the
> "protocol number".)
>
>
>     > TCP 3 was a paper specification driven by the experience gained from
>     > running TCP 2.5s.
>
> Ah, interesting. Do you concur with the position that TCP 2.5 (which is,
> apparently, undocumented - except perhaps in those minutes) was basically the
> same *protocol* as TCP 2 (which had fixed-length addresses, BTW), but with the
> *implementation* split up into separate IP and TCP layers?
>
>     > It was very shortlived (as I remember there was only one implementation
>     > ..  all the rest of us skipped ahead to 4).
>
> Most interesting...
>
>     > TCP 3's design was debugged by myriad email discussions and supplanted
>     > by TCP4 which actually got implemented by the prior 2.5 implementors.
>     > The 18-19 September 1978 meeting notes list the schedule for "TCP 4's
>     > ready for testing":
>
> Ah! I don't have those minutes, but it sounds like TCP/IP-4 was done before
> then, then...
>
>     > Prior to that, the TCP Meeting Notes of the 15-16 June 1978 meeting
>     > include:
>     > "Vint presented his goals for the meeting...
>     > The format of the TCP and INTERNET headers is to be firmly decided at
>     > this meeting ...  "
>
> Well, not to cast too wary a glance at this (potentially valuable) data, but
> no doubt the same thing had been said on other occasions... :-)
>
>     > Of course, any "decision" in that timeframe has to be viewed in the
>     > context of the philosophy of "rough consensus and running code".
>
> Also a good point.
>
>     > and later in the minutes:
>     > "The format of the headers was decided.  [See "Latest Header Formats"
>     > IEN 44.]"
>
> Indeed, IEN-44, Postel, June, 78, "Latest Header Formats"...
>
>     > This almost certainly happened on the second day - 16 June 1978.
>     >  ...
>     > So, depending on whether you're interested in the birthday of the
>     > specification or of the actual working implementation... somewhere in
>     > summer-fall 1978 would make sense to me at least.  I'd vote for 16 June
>     > 1978 - the day "the format of the headers was decided".
>
> That sounds like a good date to me, too - and I think the balance of the
> evidence is that that is indeed the day the final TCP/IP-4 headers were
> thrashed out.
>
>
>     > All of these messages/notes were emailed to
>     > [ISIE]<Postel>TCP-INTERNET.List and made available for FTP as
>     > <INTERNET-NOTEBOOK>TCP-MEETING-NOTES.TXT on ISIE.  Perhaps they live
>     > still on some moldy backup tapes somewhere...
>
> Hardcopy of all the IEN's was scanned in at some point at ISI, but they
> haven't yet made it online (even as scans). Probably easier to go that way,
> than try and find old machine-readable copies...
>
>     > as I remember, for a time there were TWO working groups - the TCP
>     > Working Group, and the Internet Working Group ..
>     > The joke was that the TCP group kept concluding at their meetings that
>     > changes were needed in the Internet design, and vice versa.  So Vint
>     > declared that they be merged, and wonder of wonders ... not long
>     > thereafter the headers merged as well.
>
> ?? But the headers are separate - are you talking about before TCP-2?
>
>
> Thanks for the reply - I think we've got it mostly worked out, now...
>
>        Noel
>
>
>