Re: Fun with RJ-45!
And you sit there wondering what possible technical reason could lead people to actually choose token ring
Perhaps the simple fact that it was faster than 10M ethernet (which was a common speed of the day) AND it would hold that speed when heavily loaded.
TR was 16Mbps IIRC, and because it was deterministic it would do 16Mbps regardless of how hard you tried to load it - all that happened if you tried to do too much was that everything had to queue a bit.
Back in those days, switched ethernet was a rarity - 10 and 100baseT were effectively just a twisted pair conversion of 10base2 (or 10base5) with all the same contention issues. If you search, you'll find that these max out (depending on the traffic mix) at about 40% of theoretical speed due to collisions, and if you try to push past that then throughput actually goes down. So with 10baseT your max throughput is around 4Mbps, with 100baseT it's around 40Mbps and if you push it then it goes down - with TR it's 16Mbps all the way regardless.
Also, planning was simpler with TR. In effect (IIRC), as long as you didn't create a cable loop then you just plug everything together and it'll work - in practical terms no limit on ring length, just on branch length as many of the hubs were active repeaters. With ethernet you had to follow the 5-4-3 rule - max 5 segments in total between any two nodes, max 4 repeaters, max of 3 segments populated (meaning that in a 10base2 or 10base5 network, some of your inter-repeater segments had to be point-point links and could not have any other nodes on them). And each segment had a max length (185m for 10base2, 500m for 10base5, 100m including patch leads for 10/100baseT) which could make large networks "interesting" to design unless you had an endless budget to allow fibre optics.
TR was, if you used the correct connectors, fairly idiot proof as well. Pull a plug out, and either the wall socket provides a loopback, or the hub at the other end detects the disconnected device and automatically bypasses that port.
There were two things that killed TR. Firstly, being a predominantly IBM system it was "reassuringly expensive" - basically nothing was cheap with TR. That alone was probably enough to kill it outside of certain markets. But also, the cable was quite bulky - less so that the '1/2" hosepipe' of 10base5 cable, but more so than the thinner 10base2, and even more so than the UTP starting to roll out for ethernet. Of course, with 10base2 or 5, you only needed one cable for multiple devices - but you had the potential for a simple fault to take out the entire network as described. And the cable was itself "reassuringly expensive".
I strongly suspect that had IBM (I assume it was IBM with the patents and licensing costs) been more astute, then TR could have won out. Until switches started replacing hubs in ethernet networks, there was really no technical reason for choosing ethernet (particularly 10/100baseT) other than cost.
And roll forward a few decades (now that make me feel old), and you can throw an ethernet network together with a pile of switches and short of having loops "it'll just work". No problems with the size of collision domains (which was the reason for the 5-4-3 rules). Kids today just don't know they're born ...