Yeah. Top-down vs. agile. Cathedral vs. bazaar. "The right thing" vs "worse is better". And, as you said, the agile/bazaar/worse-is-better camp totally out-executed the top-down/cathedral/do-it-right approach, to the point that OSI is only a theoretical model at this point, and TCP/IP is running in everything from supercomputers to washing machines.

Another one of the "well-done in theory, but left in the dust by reality" approaches: X.25.

As someone who participated at the periphery of the IETF, I resent the association with the ‘whatever goes as long as it only takes two weeks’ software planning movement. They did understand their limitations, and had a huge focus on practicality (running code), but otherwise they really did try to do the right thing. And aside from some obvious warts, I think they did an amazing job

I never said what you're resenting, nor meant to imply it.

But IETF, with their emphasis on practicality, on "rough consensus and running code", was far more on the agile end of the spectrum than OSI was. IETF would recognize a problem, have an RFC, and have working implementations before OSI had done anything. This meant that if you wanted something, your first chance to get it (often by years) was on the IETF road, not on the OSI road. Repeat that a bunch of times and the people who could benefit from new things all moved to the IETF standards.

Ha! I had worse, when I was a student (in '90s) I was taught some crazy stuff - OSI model but without a single word about the OSI protocols (like X.224) but with TCP/IP pulled over instead. So I was unironically taught that e.g. TCP is a layer 4 on the OSI model and HTTP is layer 7, etc. As if the model was still alive and relevant somehow.

They were still teaching that one decade ago, maybe never stopped.

Much the same here. I dug up this page in Wikipedia when I was reminiscing about my university's network. In my first year there was no "internet" access per se, but you could send email from the Vax VMS accounts via JANET to other sites.

Then the next (?) year there was access to the "PAD" (Packet Assembler Disassembler) to JANET, from which one could issue a command to connect to an X.25 address. I still remember the number of NSFNet relay, 00004001018057 [1], which in turn allowed you to telnet to other internet services. I think I logged into Nyx and a few other BBS-like things.

Then I did an industry year at ICL, a British computer company long-since engulfed by Fujitsu, where they were also X.25 focused for big networks and Novell Netware focused for small networks - but the web was starting to be a big deal and I couldn't get anyone to listen. I have an anecdote about downloading the SLS Linux distro via an archive re-mailer that I won't bore you with right now :)

Then, in my final year back at the university, the web seemed to explode and the X.25 stuff had more or less disappeared.

In retrospect it's amazing how quickly this changeover happened, even if it had been bulbbing away in the background for a long long time. From my perspective the WWW looked like the catalyst that pushed everything in favour of TCP/IP.

[1] If I remember rightly, the adjacent address 00004001018056 was another Vax somewhere that hosted a bunch of LaTeX stuff?

It was easy: the TCP/IP folks had BSD and it was running in the universities, and anyone who wanted to sell to them had to run BSD or else support TCP/IP too because it was the only easy way to integrate into those universities' networks.

Then people who went to school in the late 80s took that to $WORK in the early 90s, and by the mid-90s everything that wasn't TCP/IP had died. Novell's IPX, for example, was a thing still in the mid-90s, but a dying thing. Winsock for Windows 95 was the last thing that was needed to make TCP/IP rule -- after that there was just no way to do anything but TCP/IP.

It was a matter of network effects.

Same here, highschool student during 80/90's, technical magazines were full of this stuff.

Not saying that it's strictly how it has to be done but historically in OT environments the choice was more between (Ethernet) CSMA/CD and token ring (not so much Netware, it wanted to do its own thing) on a segment that IP was transiting.