[ih] IEN's as txt
jnc at mercury.lcs.mit.edu
Fri Feb 10 06:02:51 PST 2017
> From: Paul Ruizendaal
> - The specs and implementations of 1978/9 don't interoperate (I think)
> with those from 1981, not at the IP level and not at the TCP level. Why
> were the protocol numbers never bumped?
To start with, using years doesn't work well, for me at least (perhaps others'
memories work differently/better). I can remember that we did X before we did
Y, but I have no idea which years X and Y were in.
Assuming that you're speaking of the various changes to TCPv4/IPv4, the packet
formats are 'mostly' the same (modulo changes to things like addresses), which
is why they're all the same version number. As I have already stated on
another list, my belief is that, provided suitable addresse are picked, one of
those older TCPv4/IPv4 implementations probably _would_ talk to a modern
The changes in that period _had_ to be interoperable, because at that point we
were already running services over TCP/IP, and another flag day wasn't in the
cards. Well, if we'd had a major show-stopper, perhaps, but we already had
enough experience at that point to know that was not in the cards.
> Was there a sense in late 1980 that the specs were already final?
Looking at the IEN Index, to try and give myself a time reference, I think by
1980 we were into the 'interoperable change' period, so although I'm sure we
didn't think we were _done_, we did feel there would not be any more flag days
(internal to the TCP/IP world, I mean).
> I think the flag day decision must informally have been taken in the
> summer of 1981
I can't speak to that, others will have to.
> The specs were still changing as late as April
Sorry, exact which changes are you speaking of here?
Do note that the TCP/IP stack kept changing for years after this - e.g. the
subnetting in hosts change, which only became formal with RFC-1122 in October
> as Jack said that was in response to bugs creeping out of the woodwork.
I think he was speaking in generalities. You'd need to ask about specific
changes, and say 'why did change X happen'.
> What gave the comfort to declare TCP ready for use and that the changes
> were all done?
The changes clearly _weren't_ all done - and we knew that at the time. What we
did have was i) enough experience to know that it would 'sorta' work, and ii)
it was clear that running it would provide benefits.
> Or was it simply that it could not be postponed further without loss of
> credibility or some such?
I can't speak to the credibility aspect, but I do recall, when 'ordinary
network users' grumped about the trouble that was going to be caused by the
switch from NCP to TCP/IP, it was pointed out to them that DARPA wasn't paying
for the ARPANET for them to amuse themselves, it was paying for it so lessons
about networking could be learned - and the switch was to learn.
> As I understand, in November 1981 there was not a single production
> quality implementation of the April specs ready. Is that correct
Contemporary documentation is probably the best place to look for an answer to
that. It was a long time ago now, human memories have issues, etc, etc.
> were the spec's simply frozen for a while to allow/force implementations
> to catch up?
I don't recall any formal freeze, but maybe there was one. Check the meeting
notes, and things like memos that announced the forthcoming change, the
DDN Newsletters, etc.
And note that experimental changes were still happening, locally - e.g. MIT
started using subnetting, in service internally, around this time period. So
even if they was a formal freeze, stuff was still happening.
> The post flag-day part of Mike Muuss' tcp-ip digest seems to be mostly
> concerned with routing problems ... GGP and ICMP suggests that this was
> foreseen. Is that correct or are these issues unrelated?
I don't recall the issues you mention (again, detail is needed - it was a
_long_ time ago, and unless there was something special/important about a
problem - problems were a constant thing for decades - generally they will
have faded from memory), but my suspicion is that such problems were with the
_existing_ routing, and not to do with the need for new routing ideas (e.g.
the IGP/EGP split).
More information about the Internet-history