There are some technologies from not all that long ago that I'm really happy are truly dead and in no particular order coax based ethernet and token ring are amongst those.
Building those networks was doable (with the right tools), but debugging them was an absolute nightmare. Good riddance.
I had a school network that was apparently someone who decided that, since BNC T-adapters existed, there was clearly no reason to follow the 'daisychain the machines together' rule, and just spliced in a few branches here and there. Under peculiar and rare conditions, it would occasionally work, confounding the issue further.
Discovering and fixing that (surreptitiously, as a student) was a great day, since it meant we could play Doom (2?) much less blatantly since they faced away from the doors.
All running IPX and the glorious Novell Netware 3.x.
This can't be overemphasized. Both 10BASE5 and 10BASE2 were a "shared medium". Everyone connected to the same cable. It really was "an absolute nightmare" to debug.
There were no hubs and/or switches to isolate users from each other.
I don't know if they existed in the early days, but when 10Base-T got popular both hubs & switches would often have a 10Base-2 segment interface or two as well.
I'm fairly sure there was some possibility of segment isolation, because people thieving BNC terminators (seriously, they're like crack to schoolkids, along with mouse-balls) would drop off chunks of the network, but not the entire thing.
I guess it could have just been multiple NICs in the server, though.
I'm fairly sure there was some possibility of segment isolation
This discussion seems to have died off, but here's some stuff I dug up mostly from Wiki (it's been so many years since I was directly involved with any of it):
10BASE5 could be a max length of 500 meters
and have up to 100 nodes
The 10BASE5 cable must be one linear run;
T-connections are not allowed.
10BASE2 could be a max length of 185 meters
and have up to 30 nodes
So there was need for some "segment isolation" as you call it. A single Ethernet would have multiple segments (all activity fully visible on all segments), because given the restrictions on physically routing the network cables around, it wasn't possible to connect everyone in even a relatively small building to a single segment. E.g. 185 meters isn't much when it has to be snaked around the backs of desks and around the perimeters of rows of cubicles.
There were devices called "repeaters" (originally very dumb two-port amps/isolators) that would connect segments of the same Ethernet. IIRC you could even connect a 10BASE5 segment to a 10BASE2 segment.
You were limited to how many segments you could have overall, and limited in overall distance to IIRC about 1500 meters for pure 10BASE5. Every packet was visible everywhere else, it was all a single "collision domain". The distance limitation had to do with all devices needing to observe a collision within a certain amount of time (active transmitters were monitoring the wire to see if any other packet collided with theirs). CSMA/CD protocol meant all active transmitters would stop sending and "back off" an exponentially increasing random amount of time when they detected a collision.
Combine all that with "quirky" Ethernet controller chips for even more fun.
I sure don't miss any of that. Thankfully those memories have faded over time.
While not quite the same, 10 gig and 40 gig ethernet can use twin-ax connections. Big copper cabling like coax and twin-ax can carry a ton of bandwidth over short runs.
Twinax is used only for point-to-point links over short distances with manufactured cables with connectors on the end -- I understand this eliminates basically all of the pain associated with 10base5 and 10base2: It's just a somewhat inflexible cable at this point, and you're not running them in ceilings, to cubes, or such.
I still have a 3C905B handy. Those and the NE2000 were fantastic for Just Working in Linux. I think I still have some 10BASE2 "N" coax terminators and T-pieces too. All this kit was first used to play DOOM multiplayer in the mid-90s under MS-DOS using the Novell IPX drivers.
All of these technologies are now dead, except Doom.
Edit: it's kept in a box with a 56k modem, an ISA Action Replay, a Soundblaster 16, a Trident video card, and a loose Cyrix processor.
I still have a bunch somewhere. I got quite excited at the 'going for an outrageous sum on ebay' quote, but sadly they're only ~£5 each.
The NE2000's were the bane of our lanparties. Super cheap, and about as reliable as a politicians promise. They wouldn't just stop working, they'd stop everything else working.
Now that I think about it, I can probably still claim to have an SLI gaming rig (12MB Voodoo II's), if I could find something to put them in.
That must have been later on, as I seem to recall specifically avoiding 3c905 because they emphatically did not Just Work with Linux; I distinctly remember strange network behavior I didn't see with other cards. I think at the time stuff based on DEC Tulip was my choice for reliable 100Mbit Ethernet on Linux.
3c905 emphatically did work on Linux early on (as I remember this .. I was there trying to get Linux on networks ASAP back in the early days and when the 3com hardware launched I helped get the drivers written/tested on both Linux and MSDOS), but what actually happened was 3COM updated their BIOS - in spite of Linux' compatibility - due to bugs in Windows' own driver, and this kept breaking things. There was a low period where variations of the card (-B/-C series) were not supported, but this was rapidly routed around as damage, however, and 3c90x support has always been pretty good on Linux ..
Tulip support was great too, back in the early days, as well as NE2x and so on ..
Interesting, I must have just gotten in at a bad point. Certainly when I was in sysadmin in the 90's and given a 3c905(-probablyB) and told to get it to work I had very little luck.
We still had some of this in university... I remember making terminators consisting of a coax plug and a 100 ohm screw pot, we'd tune it to 50 and then move up or down depending on what the guy downstairs shouted about getting pings or not.
Well this certainly brings back (not so) fond memories of trying to figure out why the .200 net was down only to find that the intern has somehow yet again kicked apart the 10base2 tee behind his computer.
Really, really not missing physical bus topology networks.
If you remember this, then you also probably remember IBM's 4Mb token ring. Talk about proprietary. Even the cables used really weird and proprietary connectors.
Arcnet was my first network at home. Bought four cards and a passive hub for $50 from a friends father's company upgrading to ethernet. We used it for playing Doom and Duke Nukem in my basement.
Then we upgraded to 10base2 once the cards came down in price. Then finally the 4 port 10baseT hubs with a 10base2 uplink dropped in price. We had a 10 player game of Duke Nukem one night in my basement.
But up until now I have never seen a 10base5 network.
I remember helping out at my elementary school, where we had LocalTalk (RS422 with upstream and downstream cabling using special self-terminating adapters) and thinnet, with a Power Macintosh 5200 (or something similar) as a bridge in between. It ran fairly well, but troubleshooting issues on bus networks was tricky.
The even cheaper version of LocalTalk was Farallon's PhoneNet, which used cheap and available RJ11 phone cables rather than the more expensive Apple DB-3 connectors, and a terminator on each end that was basically a resistor crimped into an RJ11.
I actually ran PhoneNet over my phone lines in my apartment, since it used the two wires that aren't used by the phone. That seemed like a good idea until a thunderstorm took out my laser printer and the serial port on my Mac. Luckily, I was able to repair the Mac, but the printer was toast.
So far as I know, nobody used Apple's cabling.
A well kept secret, Apple sold an AppleTalk adapter for the PC, that worked great.
I had an injury in high school that pushed me out of PE class and into the library, where instead of sorting books I helped the librarian with the macs. The newer macs had ethernet, the older one's had AppleTalk over PhoneNet. I set up a bridge on one of the newer Macs (OpenTransport) to route TCP/IP from the Appletalk network to the ethernet network so that the older machines could get online. This meant that the school could get more computers internet access instead of waiting until they had newer machines.
So many struggles, so much fun, totally inspired me to keep building technical things. I also met my best friend to this day on this project.
So that was what the classroom networks in highschool were. The PCs had cards with the two RJ11 connectors and I remember the terminators that the troublemakers would loosen.
The backbone network and teachers network were token ring. Those connectors were huge.
That could as well be ARCNet which in addition to coax also supported single-pair Cat3 cables. From ARCnet cars I've seen I assume that pair of RJ11's was standard connector for that.
> Fast forward to 2012, and 10BASE5 is now truly a vintage technology. Anyone studying a computer science degree will at some point have been told about this stuff, because it’s very important in the history of computing.
Building those networks was doable (with the right tools), but debugging them was an absolute nightmare. Good riddance.