Back in the 90s, ninth grade me was given a two user copy of netware. I used two “network” cards I had to run a piece of bnc coax from one computer down the stairs to another computer. Nothing connected. I went down to the computer store and explained the situation to the people. It turned out I had one Ethernet and one arcnet card. They sold me a cheap Ethernet card and explained the concept of “termination” lol. That night a very excited me woke the entire house at about 2 am with an excited “it woooooooorks!!!!” And the geek networked! 😊
At university, this was late 1997, we all lived in blocks of flats on campus, 5 students to a flat. We also ran 10Base2 (ethernet over coax) from room to room, and then from one window to another in the next flat and did lan gaming. No internet, no phone lines, just quake, doom, grand theft auto 1, civilization 2, etc. Great days.
And then the letdown when you add more computers. Sure it worked for the ocasional 4 pc msdos lan party with IPX (which i had plenty), but any issue with the cable, connectors or the computers, and the whole thing would freeze. Admittedly it wasn't a big deal for a lan gaming session, most of those old games could not even continue if one player left and had to be restarted anyway. But in a working environment, like a shop running business hours, it was a PITA. Until it became star shaped, and hubs gave way to switches, Ethernet was very annoying, but at least it was an open standard just like the PC was cloned which made it popular Yes i used it in the early 90ies, netware coax, bnc, t connectors, terminators. I probably still have some around... If only arcnet was made open back then. That's like scsi vs ide at the networking level. Ethernet is so different now, where are the vampire clamps? 🙂
The reason why ArcNet was used in networked machinery (and often is still in use), is because ArcNet is a true "Deterministic" network. When he is talking about predictability, this is what he is referencing. Thus, at the network level, if you send a command to actuate a hydraulic piston on an assembly line, and the command doesn't reach the receiver, AND you don't get the response back in exactly 2 token passes... Then you know to shut down the line and not let engine blocks or rubber balls fall off the end of the conveyer belt. With NO collisions AND a way to exactly calculate the maximum round trip of a command acknowledgement (based on number of stations (limited to 256)) You could be certain either your commands were getting through or certain they were not in an predictable time. Now if you need something to happen at an exact time, no problem, hardwire it right to the PLC; but if you need something to happen (say plc to plc) within a specified time window, ArcNet is going to get the job done. Cheers!
That makes sense, most of the commands used in industrial controllers are relatively tiny. Some devices only need a couple of bits to operate but the timing and reliability needs to be perfect.
Oooh, that makes sense! In fact, thinking about it now, it's pretty obvious that's what the point of the token-passing scheme is. So I take it the machines you're talking about were running a realtime operating system? I'm curious to know which one (unless it was totally proprietary). It's funny, because I was a pretty young kid back when arcnet was a thing. I was aware of it, but that's about it. I knew about IBM token ring, but never used it. Briefly, how do token ring and arcnet differ?
@@RetroBytesUK well you have done that, with a classic British humor that I love. Trying to explain to my young techs anything older than 2010, I just send them here.
I was a Field Service Engineer for Datapoint starting in 1981 when I ended a 9 year run in the Air Force. I installed and maintained a number of large ARCNet installations in the Chicago Loop area as my first job in the IT world. Those were good times.
I remember supporting an arcnet on a steelworks they loved the half km range with active hubs and some cards. The downside was misinformed users. Some users thought they could get a speed increase by setting the address to a lower number. This kept crippling the network when they set the address to a key server. In the end we had to put metal plates over the dip switches on the pc isa cards.
Nice view from the side of the IT-Department ... Of course are those bloody users all "misinformed" and that the device they had to work with day after day (their tool!) was so effing SLOW that they got "creative" here is clearly a sign of low level "terrorism". Yeah, there is no other solution than locks, eventual punishment or putting plates over the "culprit":))) No offense, just another angle of view ... and believe me: Nothing has changed since then. I.T. kept (or doubled) its chauvinism and users are annoying[1] as usual! Hehehehe Thanks for sharing your story, Mark!:) [1] which is just a direct manifestation in how much education the employer is willing to invest into the employees (meaning to pay MONEY). You could also see this as a hidden "eff you" from your boss, as an hour long schooling could've mitigated this problem entirely without those botched solutions:)
OK, fun question: if you let users assign their own 8-bit IDs at random†, how many network nodes can you hook up before the probability of an address collision reaches 50%? †Assuming the assignment was truly random (e.g. tossing coins or dice or something), not subject to the usual unconscious human biases. #BirthdayParadox
@@lawrencedoliveiro9104 it’s an interesting puzzle, our issue was a group of users trying to set the address to a really low number. Ironically the address area where the servers and printer gateways were. Users can’t live with them… but nobody would pay the bills without them.
I did an installation of BNC coax 10 MB Ethernet Network for a local school in Poland in the mid 90s. It connected through library terminals and database server with Lantastic software. Everything worked great under MS Dos. It allowed them to run book cataloging software on 4 machines at once.
I remember Arcnet being much cheaper. The coax was a common size and the passive hubs were just a box with coax connectors and some resistors inside. You could make your own passive hub pretty easily and save some time and money. Since you could get a few PC's on a passive hub you didn't have to run a coax to every PC from the main network hub. It had some advantages.
@@JohnDlugosz One of the cool things about Arcnet was that at 2.5 Mbps it could be much looser with cabling so it allowed linear buses without a hub just like thin Ethernet, but it also allowed you to make small star networks with resistive passive hubs or large star networks with active hubs. You could also freely mix topologies within the limits of each sub-flavor. I remember encountering clusters of machines on short backbones (workgroups) which all met at an active hub in the phone closet.
7:06 Man this is so friggin' true. Ever since I was a tiny little nerd, I heard all about the great minds at Xerox parc and all the now-ubiquitous things they invented, and kept wondering why the company that made my printer isn't bigger than Microsoft.
The reason IBM and Xerox don't dominate home computing today is because the management at the time were, according to the likes of Steve Jobs, "dinosaurs who couldn't see where computers were heading". Young engineers at IBM and Xerox would come out with stuff like the mouse and the electronic pen but the management saw them as gimmicks and shelved them.
I worked for Xerox (copiers/printers) in the time frame that PARC invented all the stuff they did. If you've never heard the story, the guy who invented the laser printer at PARC nearly got fired for inventing it on his own time. The suits back east were so pin headed they could not see the future. They wasted money on insurance companies instead of bringing the future out of the lab. It has to be the most massive failure of corporate management in American history.
After all the GUI stuff with the Alto and Star, and other Xerox innovations at parc, when Xerox decided to go commercial with computers they didn't call them computers, they called them "Information Processors". Did they give their "not computer" a catchy name. Nope. They gave it a number, 820. How about a GUI, since they'd invented the thing? Heck no. The Xerox 820 series ran CP/M and was text only. The 820 series used Z-80 CPUs, had 64K RAM, and had next to nothing designed in-house. They took the open "Big Board" Z-80 CP/M design and adapted it to what Xerox wanted to sell. The final model in the 820 series was two computers in one, combining an 8088 PC with the previous Z-80 system. The nifty trick was that computer could run MS-DOS and CP/M concurrently and instantly switch the display, keyboard, floppy drives, and printer between the two operating systems, with whichever one not having access to the peripherals continuing to run.
@@Peter_S_ Yep. Steve Jobs was only a "visionary" in that he stole other peoples work. I remember when Jobs preached "you only need one mouse button", and "a stylus is not needed." And yet here we are with multi button mouses and the I-pencil. Even the I-pod and I-phone were not unique, just marketed very well.
In the late 80s we rolled out ARCNET across the office of the industrial automation company I worked for - phased out by the early 90s, but you could find at least one 93Ω BNC terminator in my desk drawer for at least a decade after that...
Yep - same here. We had a huuuuge PDP/VAX platform in 1985 and kept it going until 1998 when it all got "Y2K'd" out. We had 20 LAT networks and dedicated active ARCNET hubs all over the place. In 1998 - it all got transitioned to UTP - Cat5 - and cisco Catalyst. The VMS machines went away a few years later... and yes - I had boxes of terminators...
When QNX introduced network support, Arcnet was the only option. In fact, the network addresses got added to the command line so you could deal with files or running programs on other machines as easily as those on your own. This was practical because the addresses were short decimal numbers. I don't know how things evolved when other networks became supported.
I can also attest to the convenience of lower cost NE2000 compatible ISA NICs in the early to mid 90s. My best friend and I had a low cost thin coax network setup from my house to his running from basement to basement as our houses were close to each other. It was only possible due to NE2000 compatible NICs being less than $100 at the time. We started doing early LAN parties back then, it didn't matter if we had to buy terminators & T connectors left and right! It was so cool then to be able to play Doom or Duke3D deathmatch with more than 2 people! At it's largest we would do 4x4 matches of games like Starcraft and C&C Red Alert with 4 people at each house. We had heard of Arcnet but thin coax was so cheap at the time and easy to setup in DOS or Windows first with IPX. That setup lasted until mid 1998 when we both moved out of our parents houses and got places of our own. From then on we used regular ethernet and up until the mid 2000s has some pretty large (for us anyway) LAN parties in various places with up to a dozen or so people at a time. Those were the days!
Slight correction, modern Ethernet doesn't have collisions AT ALL. Full Duplex variants of Ethernet are only usable in switch based star topology, and that is the default as using a switch is really the only cost effective way now, and a NIC doesn't default on Half Duplex. You could force a Half Duplex connection on up to gigabit, but some NICs actually dropped support for that. Also anything higher than Gigabit Ethernet you cannot even force a Half Duplex connection if you wanted to, since it's not supported. There is no way to make a collision there
@@TheChipmunk2008 also slight correction on the gigabit, some NICs do actually support a half duplex gigabit mode, but you'd neither find a gigabit hub or be sure it will work since not all NICs have that extra mode. Anything faster than gigabit no longer has half duplex supported at all
Cool video, very well done but a small point of historical contention... there was a generation of integrated Ethernet cards based on the National DP8390/91/92 with a coax connector which predated the NE2000 from vendors including 3Com, Western Digital, and Novell itself which offered the integrated NE1000. The NE1000 in turn was really the slow (thus cheap) core of the DP8390 evaluation board with the 1986 vintage twisted pair StarLAN 1base-5 Ethernet stuff removed, no DMA, and a socket added for an EPROM, but I believe the 32 port I/O interface to the DP8390 is exactly the same. The integrated WD card retailed for $399 when Novell was offering the NE1000 for $495 but National and Novell's marketing arms worked rather closely as Novell could push NIC chip sales much harder than WD could. The NE2000 IIRC was essentially the reference schematic for the DP83901 when that came out which added DMA and I recall the DP83901 eval board schematic in the National Semi data comm handbook noting it was NE2000 compatible. I also designed a DP83901 board at the time which is why my brain is stuck with this trivia.
Ah, that history and part number helped. Was the breakout of phy from the interface card the root of the mii standard (which seems like it was after the DP8390/NE1000)?
Thanks for the extra information, you are of course correct. I did shorten the section I had on transceivers, as it felt the video was getting to be too much about Ethernet, given it was supposed to be about arcnet. I introduce the idea with the NE2000, as that's where most saw the all in 1 card for the first time, and it and its clones are what pushed down the price of ethernet.
@@ChiefBridgeFuser MII came much, much later with 100 megabit Ethernet and it's actually four 31.25 Mbps serial connections in parallel so that a 100Mbit controller chip only has to pump bits at 31.25 MHz (max) with most of it running at just under 16 Mhz which was fast for the day but still doable whereas a 125 MHz controller was out of the question. The 15 pin AUI connecter was just how everything was done at the start because the 10base-5 cable was usually in the ceiling to minimize overall length and because you couldn't just pull it through an office conduit like phone cable. When smaller networks of 10Base-2 started gaining traction you saw the backbones suddenly become departmental and venture into the cubicle or executive office and we got the pricey AUI thinNet transceivers shown in the video which were usually under the desk or in a 'wiring cubicle' which often held printers and common gear in a cubicle island. The next step was the cards with integrated PHYs, first for ThinNet and right after that you also saw pre-10Base-T AUI based transceivers for twisted pair but it was not yet the 10base-T of today and lacked the heartbeat signal. Prior to then your media choice was either coax big, or coax little. In 1990 10Base-T was ratified and suddenly actual 10Base-T to AUI adapters were common to reuse the old Ethernet cards with twisted pair. Finally in the early 90s twisted pair took off and coax became a memory but there were different PHYs for 10Base-T and 100Base-T and we started to get fiber so the makers couldn't just ditch the external PHYs just yet and it was implacticle is most cases (then) to put a dozen PHYs on a single die in a switch for example due to current consumption. By 1992 much was solved, but the ideal fab processes for a multi-PHY and for a switch IC remain different.
Ah fond memories of Arcnet. I had upgraded my store from a DOS based network to a Netware 2.2 network in 1988 and with that came Ethernet. After about 2 months of one cable getting sliced all the time at one register I ripped out the Ethernet (Thin-net 10) and put in Arcnet. Took a bit of searching to find a Micro Channel Arcnet adapter for the IBM 50Z server but I found one. Arcnet was much more robust and could work when a workstation cable was cut or unplugged. Running it was easy and the speed was never an issue. Ah the memories.... now I manage a very large 100g network with thousands of hosts, both IP4 and IPv6, sigh.
One space after period is enough. This isn't the 1500s, we have proportional typefaces now. Two is already stretching it. Yours is just taking the piss.
Arcnet is still used in some industrial niche applications. (Mostly large machines and sites where changing to something newer would be cost prohibitive.) I took part in the design of new arcnet hardware for a customer just a couple of years ago. That was a fun project, having relatively recent hardware features like PCIe and USB3 right next to your ISA bus and arcnet hybride modules. ^^
It's impressive how much old hardware is around, simply because that is cheaper than buying new industrial or medical equipment. Look up that Intel Quark developer board. A P54C Pentium running at 400 MHz, PCIe, USB2, DDR3, Wifi, it even has 16 kiB embedded SRAM
@@HappyBeezerStudios Sadly I had never an excuse to play with an Quark, but they sound interesting! Yeah, there are still products I know of manufactured to this day that use 68k (the original DIL one, not newer types) and embedded 386, often obtained through more or less sketchy parts brokers. Simply because the product has some certification that is expensive to obtain. ^^ I am not generally opposed to that as long as the parts are not salvaged ones, but I question if that still hits the spirit of those reliability oriented qualification processes...
I was the Eastern US Arcnet specilist for Datapoint. I used to carry around a Compaq "luggable" with a special Arcnet card called a "Sniffer". (thank you Dr. Harry Saul). It had a special modification that would allow me to see all data from all packets. As a specialist, I would test customer installed network. Arcnet was SO forgiving (wiring wize)... One long weekend, I forgot to bring home some coax to connect a couple machines at home. I ended up just using some lamp (zip) cord... Stripped it back to wrap around the outside of the BNC connector and then sharpend the other wire and stuck it in the center core.. It worked for what I was doing. Great memories with Gordon Peterson... BTW, a little known fact, the internal code name for ArcNet while under development was "Internet".... Crazy times.... Great presentation and thank you !
Quick story: in the early 90s we used to have mini-LAN parties and play network doom. Someone managed to get a handful of what we thought were ethernet cards. Couldnt find any DOS drivers so ended up putting out feelers on usenet (from memory). In the end, someone identified the cards as arcnet cards - that person was the inventor of arcnet, John Murphy :) We never did get them working :D
One small detail.... Coaxial Ethernet was only 2 Mbit for it's standard speed. Ethernet didn't get 10 Mbit until it switched to twisted pair cabling. But you have told a great story and told it well. And you hit the nail on the head re speed vs collissions.... When I worked in IT and supported networks the collision light on an ordinary Ethernet utp hub was ALWAYS on.... Like the collision light didn't blink... It was just on .....
I recently moved and while going through some old boxes I came across some spreadsheets where we recorded addresses for each card on a few networks I was responsible for back in the day. Ah, the memories! Passive hubs were evil - we eventually just changed them all out with active hubs; was far cheaper in the long run to just have active hubs everywhere. And I vividly remember trying to explain the collision thing to people - you are absolutely right. In the end "big number" won out because it was simpler to understand. It was so frustrating back in the day.
Also on collisions the best analogy I used was CB Radio - since they were still pretty popular in the 80's. Two people talking at once = garbage; that's a collision.
When I was first introduced to Ethernet in a Comp Sci lecture in 1979, as I recall, the lecturer had to admit that engineers were already prejudiced against the idea of such a lackadaisical approach to bandwidth management. I mean, collisions, for goshsakes! But he also said that real-world experience was already proving, to the surprise of many, that these collisions mattered a lot less than you would expect in practice. About a decade later, I remember a colleague saying that Sun Microsystems (great networking champions, as you might recall) had done a test involving two pairs of workstations on the same Ethernet wire, the members of each pair exchanging packets with its partner at maximum speed, and overall getting to 98% total utilization, even after the collisions between them. So don’t try to say that “users didn”t understand collisions”, when their own experts kept pointing out the issue at every opportunity. Ethernet won out because it worked.
@@lawrencedoliveiro9104 I don't know about Sun, but DEC did such a report in 1988: "WRL Research Report 88/4: Measured Capacity of an Ethernet: Myths and Reality," by Boggs, Mogul and Kent. And yes, for many cases (though not all) Ethernet is quite efficient even with a lot of hosts at high offered load (i.e., all transmitting data). It ranges from about 70% for very short (64 byte packets) up to somewhere in the 90s, IIRC.
I worked with arcnet. Gosh forbid you didn’t have an accurate list of all of the address dip switches for every computer. It could take forever to figure out who had the same address on your network.
@@Firthy2002 You didn't want to be the keeper of the spreadsheet, but if you didn't and other people weren't as diligent your life would be a living hell - so I was always the keeper of the blasted spreadsheet. But at least the network was stable! Also cards with the switches out the back were evil - we only bought cards where the switches were internal for reasons others have talked about elsewhere: those pesky users! Networks are always easier if you don't have pesky users running around mucking things up.
In 1986 the first LAN I installed was a nondedicated Novell 86B system using a single passive ARCnet hub. We were initially going to use GNet but we switched for a reason I don’t remember. The main purpose of the system was as a file server to share Microsoft Word (for DOS) files between the four computers. Future networks used ARCnet into the 90’s. Most of my customers were accountants and law offices. The star topology worked well with excellent response times. It wasn’t until the Ethernet switch arrived that I started believing in and pushing clients to 10BaseT. The convincer was a school district with a 10BaseT network running Windows 3.1 (might have been Workgroups) on diskless workstations booting from the network. All the classrooms involved tended to power up at the same time and it was a severe strain on the network. The server handled it just fine because Novell did their caching very well. I added a shiny 3Com 24 port switch to replace a hub for one of the rooms and it was an instant success. We never looked back.
"They also have a side hustle in producing PCBs." Whatever the sponsorship deal is, they're getting a bargain - this was one of the rare advertisements that's truly entertaining, literally LOL. Now back to watching the video....
Good watch, i remember the surge of network cards in the late 90s going from coax to 10 base-T with hubs to 100mbit switched. It seemed to happen real fast even at the time.
I remember many many years ago setting up an Arcnet system where I worked. I still have the crimp tool for the rg58 cable & bnc plugs. Oh the memories lol.
@@pctrashtalk2069 Oh yes, I've used the tool for other stuff, but remember at the time the tool costing a fortune (Around £100 from memory) and it the Mid 80's that was a lot of money lol.
@@mikebroich1487 To be honest, this was back in the middle to late 80's (nearly 40 years ago). You are probably right, but then I can't remember what I did last week lol.
One of my first clients in the early 80’s as a field engineer for a custom semiconductor company was Datapoint in San Antonio TX. I still have nightmares about trying to get someone there to commit to a custom design for their Arcnet card. No other high tech in SA at that time, so it was a long trip from Dallas to wait all day for many meetings that never happened. That company killed itself, sorry to say.
After the venture capitalist Asher Edelman purchased Datapoint he basically stripped it for profit, selling off any worthwhile technology and Assets then sold the husk to Datapont UK in June of 2000. In the UK Datapoint had specialized in Telephony systems alongside Datapoint products and they re-purposed the brand to service only this market. Moving from the old book repository building they had occupied for two decades in Neasden North London, the new Datapoint UK headquartered itself in Brentford, England and had 5,000 client sites in 41 countries worldwide. Unfortunately Datapoint (U.K.) Limited was dissolved on 14 November 2013 so the dream ended there. I worked for them in the 80's in Neasden and they were a wonderful company with a real sense of family, with many staff keeping in regular touch today. Our team was still meeting once a year right up until the Covid pandemic restrictions, which speak volumes, happy times, sadly missed.
In the 80s, Byte Magazine columnist Jerry Pournelle became interested in networking and ARCnet seemed the way to go at the time. So the cabling was run in the walls of a portion of his house, known to his readers as Chaos Manor. By the time I became a friend of the family, these mysterious coax jacks were long unused and it was years later that it was mentioned that these were the remnant of the first network in the house. I'd heard of ARCnet but Ethernet was completely dominant by the time I got interested in networking, so I didn't recognize what it was.
@@greggv8 That was true up until the point in the 80s when it came down to just having a reasonably recent PC. I built a few of those for Larry in the 90s and a bit later, until he lost interest in gaming and a good off the shelf Dell or HP made more sense for fulfilling his needs. When Jerry was first trying to get Larry onboard with word processing it was a remarkable thing to be able to say you had a computer at home.
I still use arcnet today. A line of fire alarm panels made by Honeywell use arcnet for communication locally between the boards and long distance between the panels (up to 4k feet if I remember right) on 18awg twisted pair. It works, most of the time :D
I remember those CentreCOMM MAUs from my first IT job, working for a university in the 90s. As they fell out of use they began to by used as doorstops for self-closing comm closet doors when we needed to be in and out of them often while doing work. I once found a stylish MAU under the data center floor when cleaning up the years of cables left when equipment was removed. I found an end of a length of coax and gave it a gentle pull, it eventually stopping because it was clearly attached to something on the other end. Upon following it to the other end I found a MAU in beige and light brown with a quarter round profile.
I had heard of ARCNet long ago, but never really knew anything about it other than it existing. And especially never heard of the Victor Moore/8008 (guessing that was supposed to be the 8008 rather than the 8088)/Intel bit. Incredibly fascinating and educational video, downright eye-opening, and as always your visuals are fantastic. Thanks!
Yes, should have been 8008. And Victor Poor was a self-taught tech genius, and vice-president / tech guru at Datapoint. He, and a young fellow / ham-radio guy named Harry Pyle created the 8008 architecture over Harry's Thanksgiving college break. Both have oral histories online, recorded by the Computer History Museum.
I remember mentions of ArcNet in the late 80s and early 90s relating to networking on the Amiga. Although I've never knowingly encountered an ArcNet network myself (my high school was on an Ethernet network in '95). I discovered the lack of ArcNet-Ethernet routing hardware myself not too long ago when I was looking into how someone might get an Amiga onto a modern network.
I was trying to see if cisco had done an arcnet card for any of there routers but could not fine one. I do have an arcnet card for an amiga 2000, but I could not find where I've put it.
I remember experimenting with some old ARCNet gear in the early 90s. My instructor once said "you can run ARCNet on barbed wire!" Well, I never had barbed wire, but I did manage to run it on coat hangars and a piece of broken Ethernet cable.
coax ethernet was fairly popular in some uses to late 90s. not having to set station addresses and not have switch ports for everyone was pretty useful at higher headcount scene/lan parties. sparks ahoy
When runs were longer I ran into cases where station A could talk to B, and B could talk to C at the far end of the coax, but A couldn't talk to C. Some cards handled the long runs better than others. I watched it happen with an oscilloscope... Sure did like the simple cable runs and termination though :)
At my last employer, we had a Technicians’ Department with some folks who knew their network hardware stuff. (My specialty was in software things Apple- and DEC-related, including configuring AppleTalk protocols.) One of the gadgets they had was a Time-Domain Reflectometer (aka “TDR”) which was handy for finding discontinuities in network cables.
Another excellent video. Maybe use the spinning blue network background a little less though. It's rather dizzying when your trying to look at a still image with a spinning background. Keep up the awesome videos.
In the late 1980s while I was an undergraduate at the University of New Mexico, I worked for a small company that eventually deployed a Novell Netware network using Arcnet to connect a growing number of 8088 PCs to our 80386 file server to run dBase III+ applications. It worked well for what we needed, allowing any of the workstations to access the database on the server and run independently. As a Computer Science major, I mainly had been introduced to OSI and IP networking and to a lesser extent Ethernet prior to the company installing Arcnet, but I can see the advantages, especially for a small start-up in that era.
dBase III+ now there is a blast from the past, I remember getting a NLM (maybe for a later version of dBase) and getting that installed and running that allowed for much better locking. It stopped our occasional bits of database corruption, and improved performance for us as well. Its a shame the way dBase lots most its market to Access.
The stuff CTC was doing seems amazing at the time. .. distributed storage and processing, integrated data and voice .. with micros .. in the late 70s and early 80s. We didn't get that where I work until 1997
My first adventures with a home LAN were with cheapo Realtek NE2000 compatible cards and coax cabling to the next floor so that we could do some LAN gaming. Quickly upgraded to a 10/100 hub, though, because the whole coax cable/T-piece/termination thing was very flakey, probably because I could only afford the cheap noname stuff. Upgrading later again to a 3com officeconnect switch and 3com 905TX cards was a revelation!
These videos are fascinating, I started my networking journey in the late-90s, and everything was Ethernet in the places I worked. So learning about the history of networking and different standards is immensely fascinating to watch.
I remember one factory with arcnet that had a wall that was over a foot of concret and a steamline throughit and we connected the arcnet line on each side and the damn thing worked great. That and it was the only way to get a network connection to that part of the factory much easier than trying to get a hole drilled.
@@SatumangoTheGreat yeah it was the only way short of running a cable hundreds of feet and well going through massive amounts of work my lead asked me for ideas and I said 15 muntes or so for us to test it and if it works then we save a HUGE headache customer saves a hella ton of money we all go home for the day if it doesn't we are in for at least three days work at least and even then it is going to be a major pain in the butt. We tested the signal was well within spec on the other side and it worked flawless. Definitly not a recommended way to do it but it worked =) OH and drilling a hole wasn't an option as the customer made that VERY clear.
That issue about not having to manage station addresses is a big win for businesses, especially those that didn't have dedicated IT management teams. Ethernet is simple; you just plug in a new device and, generally, it just works. Plug and play is the thing. Of course when it became just twisted pair and hubs then switches that made it all very easy. It's also a lesson that IBM didn't learn as SNA was the opposite of plug and play. It was very secure of course, but it required armies of administrators and IT department bureaucracy to set up your printer. Even in large corporations, there were semi-autonomous units that simply didn't appreciate all the delays in getting the IT department to get their new printer working on the network.
You're right not having to manage station numbers was a big win, once ethernet moved to utp the problems with its cabling randomly breaking and you having to follow the cable around the buidling went away, which then made ethernet the less hassle option.
For a long time, SNA wasn’t even a “network” as we would understand it today--namely, the whole peer-to-peer element was completely missing until a later version of SNA came along sometime in the 1980s to play catch-up to TCP/IP, AppleTalk and all the rest. The new version was identified by acronyms like “LU 6.2”, “PU 2.1” and “APPN” (“Application Program-to-Program Networking”, if I recall).
@@lawrencedoliveiro9104 You have to love IBM and their acronyms/abbreviations. However, I have a soft spot for the mainframe architecture as I started out programming using IBM 370 assembler code writing operating system code (for a non-IBM OS as it happens - a long story). As far as SNA goes, and other things IBM, I recall a lot of what might even be called denialism on the part of the IBM technical support community when open systems came along. So many of them had gone through the way that IBM trained system programmers, DBAs and the like that they'd almost been indoctrinated into thinking that the way IBM did things was the only way. Their world was being turned upside down by what looked like chaos. Even when they worked on flawed things like CICS, CKD architecture disks, and the wonderful complexities of VSAM.
I remember Arcnet! Ahh good times supporting a site with an 'eclectic' mix of kit. The organisation (a very large, rich one) bought stuff on Ebay to add stations or for spares. Every so often the couriers would bring in job lots of assorted PCBS, dumb terminals, chassis etc in boxes and it was a case of "Make something work".
ArcNet lived on as the basis for a number of industrial control system network protocols - the reliable timing and intrinsic confirmations made it very suitable for distributed control systems.
Nice. I vaguely remember going through some arcnet-training at Xerox 23 years ago, together with Novell netware and some other things I never had any use for.
I got my Certified Novell Administrator certificate for Netware 3.x, just in time for the debut of Netware 4.x. I just bought the big Netware 3.12 book, read it cover to cover, then went to the local Sylvan testing center and passed the test. Wasn't long after that Sylvan and Prometric merged to become Sylvan-Prometric and as most monopolies do, they massively raised the pricing for their Netware and other certification tests. I *was* going to buy the big book for Netware 4.x and get that CNA cert but when I saw the price for the test I said hell with that. It would have been a waste because Netware ended up being a dead end.
I remember faffing about with hardware in the 286 days and finding what I thought were thin-net cards as they had a coax connector and no AUI port. I had always assumed the dip switches were for setting IRQs/Ports. But after seeing this, I guess they were ARCNET
My first "real" job was for my hometown school district. Novell network and tons of arcnet. Even as other portions switched over to 10base2 or 10baseT longer runs between buildings (grounding issues aside) were only possible with arcnet. That worked until budget was finally available to make those runs fiber. :-)
I remember getting our first fibre link, it felt so big budget at the time. We had to get due to grounding issues between different parts of the building, there was a 50v potential difference, which was sufficient to kill equipment, and hurt.
@@RetroBytesUK a friend used to do internal IT support for Albertson's (one of the largest grocery store companies in the USA). At one location they had a gas station in the parking lot. The system in that small building connected to the server in the main store using an Ethernet cable run through conduit. Periodically they'd get a lightning strike in the center of the parking lot, right on the network cable. Various computer equipment at one or both ends would get fried. (IIRC they'd also have to pull new cable through the conduit.) After a few times of that the company upgraded the link to fiber optic and never had another lightning strike in the parking lot.
Arcnet was my career for a few years and it is installed in some very interesting and quite secret facilities. Such a shame they did not do a much faster version but this is why having one corporation with a monopoly is a very bad idea.
Didn't realize that the phy interface to the mac was external to the card on early implementations. Helps explain why mii and rmii are part of the 802.x standards.
I shortened the bit about how the transceiver started to get included on the card, and just moved to the NE2000 as that was the first card with it included most places first used. There where other cards that did it first including the NE1000, but the NE2000 (which was basically the NE1000 with DMA) was the one that sold in volume where most got to encounter it. Someone in the comment has given a much more detailed history of that bit than I did, as I was trying to not make the video too much about ethernet.
@@RetroBytesUK Ah, you've sucked me into looking at this ancient history. Looks like that wasn't the mii interface. Seems like it was probably the inspiration for it. Motivation to separate the analog parts of the phy from the purely digital functions of the mac.
Arcnet was my comms of choice for ROV comms - I could get 2000ft of communications to the sea bed in the North sea on a twisted pair working at a depth of generally less than 1000ft - then it was picked up by Tritech for it's stability (token passing) for their sonars and profilers etc. good engineering gets selected by good engineers, not accountants, they listen to sales lies and crap. However 10 Gbit and faster rates have rendered this moot on fibre optic systems. these are reaching more than 11km the bottom of then Marianas trench (approx. 10,880 metres). Standard Moog kit 907 types guarantee 10Km reliable comms now.
I serviced some businesses in the late 90s and early 2000s with these lol, I learned about it while in the field as a tech for a large warranty service company, don;t really miss that.
I once worked at a company where we have Novell servers with both Arcnet and Ethernet cards in them because the back-of-house was all on Arcnet where as the front was Ethernet (10-Base-T coax), so the file and print servers (back when those were separate things) had to be able to talk to both networks. We also had dumb terminals for a Unix system that where connected by serial cables to concentrators where said Unix system was connected to the Ethernet (mostly so it could use the Novell print server). There was also a small Token Ring setup in one of the test labs, so we had it all. Ah, fun times.
FDDI was a very interesting technology, I only got to see it deployed twice. Once on a site that used token ring for the clients and an FDDI ring for the servers. The other was a cable company that used it to link their street cabinates prior to using docsis, it was the backhawl for all their telephony services.
ปีที่แล้ว +4
Nice one! Thanks. Learnet quite a bit about ArcNet. One thing, though: you mentioned Token Ring having 8 or 16 MBit. But: Token Ring had (on the PC) typically 4 or 16 MBit/s. There as been other TokenRing implementations, too, e.g. 12 MBit with Apollo Domain workstations. But the beforementioned 8 MBit/s Token Ring is either quite an oddity or not existing :)
OMG. Another Domain-iac on a TH-cam. Too bad I can only give one thumbs up. We had a giant Apollo token ring network with the office split into multiple sub-rings that could be switched out/in in the wiring closet when we needed to add connectors to the sub rings
ปีที่แล้ว
@@ppokorny99 I never experienced Apollo Domains in their natural habitate and time context (it’s not my fault! I’m born too late!) but only years later as nice workstations. The 2000 and 2500 now serve in our Regional Retro Data Centre, alongside RS/6000, Acorn Archimedes, Apple III, Commodore 8032, Atari Mega ST and many others. Coming back to your point: I still need a way to route packets between Apollo TR 12 MBit and the other network systems we operate (FDDI, 4/16 Token Ring, 10BASE-T1L, 100VG-AnyLAN, IBM BROADBAND, LocalTalk, FlashTalk …) What was the main purpose of your Apollos?
Came here to say this, comments did not disappoint. I have put in brand new 4/16 mbit real IBM microchannel cards back in the day, so quite familiar...
I'm always into videos that start with ripping Intel a new one... :) Also, I remember, at the mid-to-late-90s, I was still running into ARCNet networks as a computer-wiz-kid, but there was very little info available (to me), and I already had some Ethernet knowledge, so the troubleshooting was usually just to rip the thing out, and replace with the by-then dirt cheap Ethernet. But I was always intrigued by it, like some mystical stuff that was built by "ancients". So today, after all these years, I learned something. Thanks!
That video brought back many memories. I worked for a major financial institution in Canada who inbitially had a large Datapoint deployment; both with true Datapoint hardware and third-party servers from Performance Technology,. Then a Banyan / VINES network was added in to the mix, all running over ARCNET. The joys of managing the network address on each PC and of ensuring we didn't exceed the max number of stations on a single ARCNET network. Some areas did start to use ethernet, and of course there was also Token Ring.
The problem with the throughput topic is that while the theoretical “worst case” throughput of 2.5Mb ARCNet is higher than 10Mb Ethernet, the conditions under which that would happen have NEVER been demonstrated in the real-world. Collisions are uncommon in any demand network below 50% utilization, which means that until you see average demand on a “thin-net” exceeding 5Mb/s for an extended time-period, you’re not going to see any degradation. At that point, you’re already at 2x the hypothetical performance of ARCNet, and you don’t reach saturation until 70%-90% in real-world situations, so the real-world performance of 10Mb Ethernet is between 2x and 3.5x the maximum throughput of ARCNet. IBM’s Token Ring suffers from the same limitations, even in 16Mb version. In 1988/1989, I did real-world testing of 10Mb Ethernet (thin-net) vs 16Mb Token Ring, complete with MAUs, on 286/386 workstations and a 386 based server. In every single test, single station load and many station load, each station attempting to transfer as much data as possible, the 10Mb Ethernet outperformed the 16Mb Token ring in both single station and composite multi-station throughput. And Ethernet was significantly cheaper too. 100Mb Ethernet then made Token Ring completely irrelevant.
Actually, the (incorrect) figure that most people pull out of the theoretical studies and modelling studies is that Ethernet tops out at 37% offered load, which at 3.7 Mbps is already 50% higher than ARCNet's maximum speed of 2.5 Mbps. But that 37% figure doesn't actually bear much, if any, relation to how real Ethernet works. If you want all the gory details, the the 1984 paper Boggs, Mogul and Kent, "WRL Research Report 88/4: Measured Capacity of an Ethernet: Myths and Reality," is a good read. Note that the saturation at 70% figure they measured in their study was with 64 byte packets, which is unrealistically low for almost any situation.
Great explanation of ARC Net. I still have here a few 3Comm ARRnet/Ethernet ISA cards. Tucked away in my archive of "old DOS" stuff is a version of Netware Lite that you have to load the protocol drivers in the correct order to get it to work. Was fun.
My first job in computing was working for mintex the brake and clutch manufacturer. we had sales depots around the UK each was equipped with datapoint computers recording sales made during the day and generating invoices. My task was to use dial up technology to obtain the sales records, run a programm to update stock records which then went on to generate restocking levels and factory demand. Most of the 23 depots had 3 floppy drives on the computers where the head office where I was based had Winchester hard drives of around 10 mb to store the information. Backups, report generation etc. When the system went wrong usually in end of day updates or data communications I was the one who remotely working in hexadécimal to write end of records and close files. 5 years experience lead me into teaching and I never looked back. Thanks datapoint and mintex and of course dos. Happy days
I worked with both Datapoint and Netware. 3Com introduced ethernet cards to offload the network processing to the card such as 3C503 and the server 3C505.
Yup, the company I worked for in the late 80s pulled out our Arcnet for 10Mb BASE-T around 1989 when we brought in a new SCO/Oracle system running on a Tricord and never looked back. NetWare286, modem banks, Telebit dial-up routers, ODBC and Lotus 123... Fun times.
I wired one school with the ArcNet+ network in 1992. As college had taught only very little outside Ethernet, I was first a bit puzzled as to what it was. However, crimping coax cables was a breeze after a handful of failures.
Lived through this running a computer network for my university from 88 - 92. With our 8086 and 80286 labs, they were all ARCnet. By the time the 386 and 486 machines were coming around everything was Ethernet. Novell made a “MPR” multi-protocol-router that would bridge our ARCnet to Ethernet (IPX -> TCP was our use case). Price was the biggest factor driving the convergence to Ethernet. The other driving factor was the emerging internet and TCP/IP. ARCNet’s limited TCP/IP stack support in most PC OS’s was a huge drawback. Then you add Windows NT server and Linux starting on the scene and things get really grim for Novell and by extension ARCnet.
You didn't mention that the big reason there weren't lots of companies making arcnet cards was because Datapoint required every manufacturer to buy those proprietary black coated PCB modules from them. That's the rectangular piece close to the BNC at 12:00 As a single source supplier for a must have component, Datapoint could pretty much set the price that arcnet cards would sell for. The 1989 introduction of the Ethernet switch by Kalpana rang the doom bell for most other network technologies. Switches could be swapped in place of hubs and instantly boost aggregate throughput on a highly loaded Ethernet network, while requiring no reconfiguration at any of the connected computers or other devices.
I had one large pre-installed arcnet system that was upgraded to 10 mbps Ethernet. To bad it was before twisted pair was used. So we used coax 58ohm? (to many years ago.) Also had one token ring network existing. Added a 2nd card in the Novell server to expand with Ethernet. But, that was the life when I started into computers. I found out later on Arcnet was already refereed to as pcnet. Then I found a few other cool things about it. Like radio shack used it for a trs-80 network.
@@gorillaau Ah, to many years ago. My fav was them unplugging x amount of computers and having it go down. So happy when Twisted Pair and hubs came around.
The 8086 is pretty district architecture from the 8080. Although you could argue that it's a combination of ideas from the 8080 and the 6502/6800 processors, so not entirely unique.
Also despite what you may have heard, the 8086 is not source compatible with the 8080. This is pretty obvious when you think about it as the 8080 has a bunch of instructions that the 8086 doesn't and they have different register names. I believe this misconception comes from the existence of programs to convert code from one to the other. Over time "You can convert between then" becomes "They are compatible".
Datapoint, now there's a blast from the past. I started my IT career back in the '80s working as a programmer for a UK software company writing in Databus - those were fun times.
I encountered Arcnet in an embedded application in about 2002 I think. At that time it had two incompatible physical implementations, one using what I think was the original AC transformer coupled implementation and a second one using RS485 trancievers. Unfortunately this application used the RS485 variant and the link shared a cable with a 56V DC supply, so the trancievers died frequently. I was asked to repair the boards so they could go back into service. My first "fix" involved fitting Maxim "fault protected" 485 trancievers which could withstand the full 56V they'd see in a worst-case misconnection. So I thought ... then I'm presented with units that will communicate with the host PC directly, but not over the long cable. Turns out the terminating resistor has ceased to exist. So now I fit a PTC thermistor with a cold resistance near enough to 100 ohms. Meanwhile a newer generation of units use the AC coupled version and later a variant with a DC-blocking capacitor so a misconnection can't burn out the transformer. As an experiment I tried adapting their two-pair cable to 100baseTX and I could run 100baseTX over it and with a bastard POE implementation could have run the DC over the same cable with half the voltage drop of the original system.
I remember having to use a vampire tap to connect a computer to a thicknet cable. The cable was marked to show you where you could put a tap. When you were having problems with thinnet, it was usually because you were missing a terminating resistor at the end of a line. People would move desks and computers and weren't paying attention to the cables. This was in the early to mid 90's.
I was always jealous of the NE2000 cards because my company ran Banyan Vines servers. Now THERE is a topic for a future video. Vines required special cards at first. It wasn't until the 3Com 3C503 cards appeared that we got a cheaper option.
I never got to see vines in its special card phase only after you could use etherent. Some times I feel like the only person who remebers it, as I get very blank looks when ever I bring it up among old networking types.
We used ethernet. The cards just had to be Vines compatible. Looks like they added NE2000 and TCIP/IP support with the last versions.We had moved to NT by then. For a brief period, Vines was the only NOS that could scale REALLY big.
Microsoft hired Jim Allchin away from Banyan to bring StreetTalk to LAN Manager. A couple of years later we got Active Directory. So in some ways it is still around.
Year 2000 (Y2K) retired the Netware 286 systems I supported. There were a few Netware 4 systems that had to have hardware die to get them to their heavenly reward.
It was my job at CompuPro, in the 80's, to make the Arcnet adapters and software for both Pc's and the S-100 machines. The most difficult thing was to make the state machines on each node work, when the protocol was not a symmetrical dialog process, as is the case with Ethernet. That meant that the token bus approach was inherently less robust against signal dropouts or packet collisions.
So, the possible event points were: reconfigure after reset, waiting for token, getting token, releasing token, rinse repeat. Or, get token, send packet, wait for ack to sender, get ack from recipient, wait for packet reply from recipient, send back of reply from sender to recipient, rinse repeat. At each of these points a reconfig event might arrive asynchronously, requiring resetting the sending or receiving packet counters kept in each node, except for the times when it doesn't. And theassymetry occurs at the time interval between the first ask and the packet reply. There is where the most danger lies for losing sync of the packet counters.
ARCNET also supported a loop through cabling topology using BNC Tee connectors just like Ethernet on Thinnet. Exactly the same except that ARCNET used RG62 coax cable at 93ohms and Thinnet used RG58 coax at 50ohms. Hubs were not required for Arcnet but they did did make cabling easier and the network reliability was also higher as the loop topology is easily broken at any attached computer taking the whole network down. Note too that you could also have both hubs and looped segments with ARCNET. I was a big user of ARCNET in the 1980s and 90s for industrial computer automation networks. I even designed and fabricated custom ARCNET embedded hardware. I miss those days!
In 90's I participated in a project that incorporated Arcnet into an embedded network. Included was also a proprietary ISA-card for connecting a PC into the system. Those were the days.
At the time, Intel never claimed that they had invented the microprocessor. In fact, they didn't even attempt to patent it, but said instead that there's nothing patentable in a microprocessor, since for something to be patentable, it needs to have a non-obvious original idea and the idea of a microprocessor was obvious to anyone working in computing at the time.
I love this channel. I vaguely remember my mom messing with some Arcnet stuff when I was a kid. And I don't mean she was in IT, just she did work on machines hooked to an Arcnet network. I remember her working with and around all kinds of stuff that is rare now or was rare even back then. Or was once popular and died off because something else took its place.
Arcnet was the initial choice for Tandy's first networking forays (TRS-80 Model 2 and 16). When I worked there, it was so slow (even for the times) that internally we called it ArcNOT and Bark-net
What a deep subject I never knew I wanted to care about 😂 Any thoughts on making a video covering Unix in general? What it is/why it’s important/how it came to dominate early on/etc?
My best friend in higschool and I ran a two-node Renegade BBS on a 486SX-25 and a 386DX-33 (I believe), and networked the two nodes via Arcnet. Didn't have the right connectors, so we used coat hangers for a while...
I worked for a company that was supposed to be focused on error proofing everything. That’s what they said. What they did was connect both Ethernet and arcnet to the same operator station. The cables looked identical, but had different impedances. The terminators had the resistance stamped on them, but they were hard to read, so they were often installed on the wrong network. Bottom line, the systems never worked out of the box.
Great video. And one thing that wasn't mentioned as a nail in the coffin is the introduction of the Ethernet switch. It mitigated the impact of collisions in many scenarios.
I'd like to see a RetroBytes video on PCMCIA and CardBus, and how it worked in DOS + Windows 3.x. Getting it to work in DOS required a lot of drivers. First there was the Card and Socket Services driver specific to the PCMCIA chip. Then for the cards there could be either Point Drivers or Class Drivers. A Point Driver worked with one specific card, for example a SCSI controller, sound card, or other device that had unique properties. A Class Driver worked with many devices of the same type, like battery backed DRAM cards or NE2000 compatible Ethernet cards. The real trick was getting the C&SS driver plus Point and Class drivers for more than two cards all crammed in simultaneously (no dynamic loading and unloading of drivers) while also having all the DOS drivers for everything else in the laptop AND still having enough free RAM to run Windows or a DOS program. But with it working one could hot swap the cards that had drivers installed. If you needed more than 3 or 4 different cards needing Point drivers you had to setup a multiple choice DOS boot to select different sets of drivers. I don't know how well Windows NT4 handled PCMCIA but I do know that Windows NT 3.5 didn't get along with it on at least one laptop. In the late 90's a friend got an old laptop and just for fun he installed NT 3.5 on it and tried for quite a while to get the PCMCIA slots to work. Nope, wasn't happening. Windows 95 erased all those limitations. It needed no DOS help for PCMCIA, though without DOS drivers loaded DOS software couldn't use the PCMCIA slots. IIRC the laptop I had with Windows 95, I had 4 or 5 different PCMCIA cards I used in it without a problem. It also had easily swappable hard drives so I did one with a stripped down Win95 install with Basilisk II as the default shell so I had a Dell Powerbook that booted into Mac OS 8.1. Since the Mac emulator was the shell, shutting it down or restarting it would make Windows shut down or restart. It was mostly a "Ain't this neat?" deal because it was a 75 Mhz Pentium with passive matrix LCD.
Don't know which video was the first one that came into my feed a few days ago and had you taking me down memory lane but thanks. I got my first micro in 1983 when I was 14, idiots at school called my dad because I wrote a paper on computers and the staff thought I plagiarized it. It's funny watching history and documentary stuff and being like "I remember that" and "I was screwing around there". In a way it might be a bit like how some of the engineers and workers on the 1950s/60s rocket assembly teams felt. Even got to do some work on a mainframe before Y2K, the only people making a fuss about that was the press. Everyone in the field know it was just a quick cash grab because the pres wanted to turn it into something they could hype to up their ad rates.
This misses the whole effect of IP which was used on most ethernet but not ARCnet. Ethernet allows for use in a WAN while ARCnet is really LAN only. IP was really the driving force which is why ARCnet is only seen on factories. Thick wire and thin wire are also confused here. Interesting overall however.
![PASULOL รามเกียรติ์ ตอนที่ 4 นนทกพบรัก [Ramakien Ep.4: Nonthok in love]](http://i.ytimg.com/vi/tMUQgmZzf_A/mqdefault.jpg)
![[TH] 2024 PMGC League | Survival Stage วันที่ 1 | PUBG MOBILE Global Championship](http://i.ytimg.com/vi/FVFT-PFwU80/mqdefault.jpg)
Back in the 90s, ninth grade me was given a two user copy of netware. I used two “network” cards I had to run a piece of bnc coax from one computer down the stairs to another computer. Nothing connected. I went down to the computer store and explained the situation to the people. It turned out I had one Ethernet and one arcnet card. They sold me a cheap Ethernet card and explained the concept of “termination” lol. That night a very excited me woke the entire house at about 2 am with an excited “it woooooooorks!!!!” And the geek networked! 😊
At university, this was late 1997, we all lived in blocks of flats on campus, 5 students to a flat.
We also ran 10Base2 (ethernet over coax) from room to room, and then from one window to another in the next flat and did lan gaming.
No internet, no phone lines, just quake, doom, grand theft auto 1, civilization 2, etc.
Great days.
😂me the same. C'mon there must be thousands out there who experienced this eureka nerd thrill...
And then the letdown when you add more computers. Sure it worked for the ocasional 4 pc msdos lan party with IPX (which i had plenty), but any issue with the cable, connectors or the computers, and the whole thing would freeze.
Admittedly it wasn't a big deal for a lan gaming session, most of those old games could not even continue if one player left and had to be restarted anyway. But in a working environment, like a shop running business hours, it was a PITA. Until it became star shaped, and hubs gave way to switches, Ethernet was very annoying, but at least it was an open standard just like the PC was cloned which made it popular
Yes i used it in the early 90ies, netware coax, bnc, t connectors, terminators. I probably still have some around... If only arcnet was made open back then. That's like scsi vs ide at the networking level. Ethernet is so different now, where are the vampire clamps? 🙂
I don't think we used IPX.
I'm sure we used IP, with Windows 95.
@@MostlyPennyCat DOOM would have been IPX.
The reason why ArcNet was used in networked machinery (and often is still in use), is because ArcNet is a true "Deterministic" network. When he is talking about predictability, this is what he is referencing. Thus, at the network level, if you send a command to actuate a hydraulic piston on an assembly line, and the command doesn't reach the receiver, AND you don't get the response back in exactly 2 token passes... Then you know to shut down the line and not let engine blocks or rubber balls fall off the end of the conveyer belt. With NO collisions AND a way to exactly calculate the maximum round trip of a command acknowledgement (based on number of stations (limited to 256)) You could be certain either your commands were getting through or certain they were not in an predictable time. Now if you need something to happen at an exact time, no problem, hardwire it right to the PLC; but if you need something to happen (say plc to plc) within a specified time window, ArcNet is going to get the job done. Cheers!
That makes sense, most of the commands used in industrial controllers are relatively tiny. Some devices only need a couple of bits to operate but the timing and reliability needs to be perfect.
That's a perfect job for a CAN.
Oooh, that makes sense! In fact, thinking about it now, it's pretty obvious that's what the point of the token-passing scheme is. So I take it the machines you're talking about were running a realtime operating system? I'm curious to know which one (unless it was totally proprietary).
It's funny, because I was a pretty young kid back when arcnet was a thing. I was aware of it, but that's about it. I knew about IBM token ring, but never used it. Briefly, how do token ring and arcnet differ?
@@LMB222 yeah a can of beer can resolve most problems. Lol
As an IT professional I love how you videos show how we got to where we are now.
Thanks, that was what as big part of what I hoped to achieve with these things.
@@RetroBytesUK well you have done that, with a classic British humor that I love. Trying to explain to my young techs anything older than 2010, I just send them here.
I was a Field Service Engineer for Datapoint starting in 1981 when I ended a 9 year run in the Air Force. I installed and maintained a number of large ARCNet installations in the Chicago Loop area as my first job in the IT world. Those were good times.
And you worked with the one and only, Wayne Pan !
@@garymiyakawa Wayne was just before my time but I knew of him. There sure were a lot of memorable people that worked the Chicago area.
I was a systems engineer in the 90s and I used to run legacy ARCnet LANs hosting Novell with IPX.
I remember supporting an arcnet on a steelworks they loved the half km range with active hubs and some cards. The downside was misinformed users. Some users thought they could get a speed increase by setting the address to a lower number. This kept crippling the network when they set the address to a key server. In the end we had to put metal plates over the dip switches on the pc isa cards.
I do know places that packed blue tack in there to stop users changing the values.
Some things are just more of an HR problem, and not an IT one....
Nice view from the side of the IT-Department ...
Of course are those bloody users all "misinformed" and that the device they had to work with day after day (their tool!) was so effing SLOW that they got "creative" here is clearly a sign of low level "terrorism". Yeah, there is no other solution than locks, eventual punishment or putting plates over the "culprit":)))
No offense, just another angle of view ... and believe me: Nothing has changed since then. I.T. kept (or doubled) its chauvinism and users are annoying[1] as usual! Hehehehe
Thanks for sharing your story, Mark!:)
[1] which is just a direct manifestation in how much education the employer is willing to invest into the employees (meaning to pay MONEY). You could also see this as a hidden "eff you" from your boss, as an hour long schooling could've mitigated this problem entirely without those botched solutions:)
OK, fun question: if you let users assign their own 8-bit IDs at random†, how many network nodes can you hook up before the probability of an address collision reaches 50%?
†Assuming the assignment was truly random (e.g. tossing coins or dice or something), not subject to the usual unconscious human biases.
#BirthdayParadox
@@lawrencedoliveiro9104 it’s an interesting puzzle, our issue was a group of users trying to set the address to a really low number. Ironically the address area where the servers and printer gateways were. Users can’t live with them… but nobody would pay the bills without them.
I did an installation of BNC coax 10 MB Ethernet Network for a local school in Poland in the mid 90s. It connected through library terminals and database server with Lantastic software. Everything worked great under MS Dos. It allowed them to run book cataloging software on 4 machines at once.
I remember Arcnet being much cheaper. The coax was a common size and the passive hubs were just a box with coax connectors and some resistors inside. You could make your own passive hub pretty easily and save some time and money. Since you could get a few PC's on a passive hub you didn't have to run a coax to every PC from the main network hub. It had some advantages.
One of the advantages was much longer range than even coax Ethernet, and that had longer range than twisted pair Ethernet.
I vaguely remember, when the price was coming down on cards, there were also cards that permitted a two-computer network without having to buy a hub.
@@JohnDlugosz One of the cool things about Arcnet was that at 2.5 Mbps it could be much looser with cabling so it allowed linear buses without a hub just like thin Ethernet, but it also allowed you to make small star networks with resistive passive hubs or large star networks with active hubs. You could also freely mix topologies within the limits of each sub-flavor. I remember encountering clusters of machines on short backbones (workgroups) which all met at an active hub in the phone closet.
7:06 Man this is so friggin' true. Ever since I was a tiny little nerd, I heard all about the great minds at Xerox parc and all the now-ubiquitous things they invented, and kept wondering why the company that made my printer isn't bigger than Microsoft.
Steve Jobs wondered the same thing and set out to do what Xerox should have with some of that tech. In the end you can sum it to "corporate focus".
The reason IBM and Xerox don't dominate home computing today is because the management at the time were, according to the likes of Steve Jobs, "dinosaurs who couldn't see where computers were heading". Young engineers at IBM and Xerox would come out with stuff like the mouse and the electronic pen but the management saw them as gimmicks and shelved them.
I worked for Xerox (copiers/printers) in the time frame that PARC invented all the stuff they did. If you've never heard the story, the guy who invented the laser printer at PARC nearly got fired for inventing it on his own time. The suits back east were so pin headed they could not see the future. They wasted money on insurance companies instead of bringing the future out of the lab. It has to be the most massive failure of corporate management in American history.
After all the GUI stuff with the Alto and Star, and other Xerox innovations at parc, when Xerox decided to go commercial with computers they didn't call them computers, they called them "Information Processors". Did they give their "not computer" a catchy name. Nope. They gave it a number, 820. How about a GUI, since they'd invented the thing? Heck no. The Xerox 820 series ran CP/M and was text only. The 820 series used Z-80 CPUs, had 64K RAM, and had next to nothing designed in-house. They took the open "Big Board" Z-80 CP/M design and adapted it to what Xerox wanted to sell. The final model in the 820 series was two computers in one, combining an 8088 PC with the previous Z-80 system. The nifty trick was that computer could run MS-DOS and CP/M concurrently and instantly switch the display, keyboard, floppy drives, and printer between the two operating systems, with whichever one not having access to the peripherals continuing to run.
@@Peter_S_ Yep. Steve Jobs was only a "visionary" in that he stole other peoples work. I remember when Jobs preached "you only need one mouse button", and "a stylus is not needed." And yet here we are with multi button mouses and the I-pencil. Even the I-pod and I-phone were not unique, just marketed very well.
In the late 80s we rolled out ARCNET across the office of the industrial automation company I worked for - phased out by the early 90s, but you could find at least one 93Ω BNC terminator in my desk drawer for at least a decade after that...
Yep - same here. We had a huuuuge PDP/VAX platform in 1985 and kept it going until 1998 when it all got "Y2K'd" out. We had 20 LAT networks and dedicated active ARCNET hubs all over the place. In 1998 - it all got transitioned to UTP - Cat5 - and cisco Catalyst. The VMS machines went away a few years later... and yes - I had boxes of terminators...
When QNX introduced network support, Arcnet was the only option. In fact, the network addresses got added to the command line so you could deal with files or running programs on other machines as easily as those on your own. This was practical because the addresses were short decimal numbers. I don't know how things evolved when other networks became supported.
I can also attest to the convenience of lower cost NE2000 compatible ISA NICs in the early to mid 90s. My best friend and I had a low cost thin coax network setup from my house to his running from basement to basement as our houses were close to each other. It was only possible due to NE2000 compatible NICs being less than $100 at the time. We started doing early LAN parties back then, it didn't matter if we had to buy terminators & T connectors left and right! It was so cool then to be able to play Doom or Duke3D deathmatch with more than 2 people! At it's largest we would do 4x4 matches of games like Starcraft and C&C Red Alert with 4 people at each house.
We had heard of Arcnet but thin coax was so cheap at the time and easy to setup in DOS or Windows first with IPX. That setup lasted until mid 1998 when we both moved out of our parents houses and got places of our own. From then on we used regular ethernet and up until the mid 2000s has some pretty large (for us anyway) LAN parties in various places with up to a dozen or so people at a time. Those were the days!
Slight correction, modern Ethernet doesn't have collisions AT ALL. Full Duplex variants of Ethernet are only usable in switch based star topology, and that is the default as using a switch is really the only cost effective way now, and a NIC doesn't default on Half Duplex. You could force a Half Duplex connection on up to gigabit, but some NICs actually dropped support for that. Also anything higher than Gigabit Ethernet you cannot even force a Half Duplex connection if you wanted to, since it's not supported. There is no way to make a collision there
didn't know that about gigabit... but makes sense, switches are cheaper than hubs now, if you can find a hub
I still have a 10baseT hub on my network... it supplies my chromecast and the AP for my guest wifi. I use it to slow both of em down ;)
@@TheChipmunk2008 oh that's funny, still most likely full duplex so no collisions
@@TheChipmunk2008 is 10 mbit enough to stream video though?
@@TheChipmunk2008 also slight correction on the gigabit, some NICs do actually support a half duplex gigabit mode, but you'd neither find a gigabit hub or be sure it will work since not all NICs have that extra mode. Anything faster than gigabit no longer has half duplex supported at all
Cool video, very well done but a small point of historical contention... there was a generation of integrated Ethernet cards based on the National DP8390/91/92 with a coax connector which predated the NE2000 from vendors including 3Com, Western Digital, and Novell itself which offered the integrated NE1000. The NE1000 in turn was really the slow (thus cheap) core of the DP8390 evaluation board with the 1986 vintage twisted pair StarLAN 1base-5 Ethernet stuff removed, no DMA, and a socket added for an EPROM, but I believe the 32 port I/O interface to the DP8390 is exactly the same. The integrated WD card retailed for $399 when Novell was offering the NE1000 for $495 but National and Novell's marketing arms worked rather closely as Novell could push NIC chip sales much harder than WD could. The NE2000 IIRC was essentially the reference schematic for the DP83901 when that came out which added DMA and I recall the DP83901 eval board schematic in the National Semi data comm handbook noting it was NE2000 compatible. I also designed a DP83901 board at the time which is why my brain is stuck with this trivia.
Ah, that history and part number helped. Was the breakout of phy from the interface card the root of the mii standard (which seems like it was after the DP8390/NE1000)?
Thanks for the extra information, you are of course correct. I did shorten the section I had on transceivers, as it felt the video was getting to be too much about Ethernet, given it was supposed to be about arcnet. I introduce the idea with the NE2000, as that's where most saw the all in 1 card for the first time, and it and its clones are what pushed down the price of ethernet.
Also nice to get it from someone who designed a board for the DP83901.
@@ChiefBridgeFuser MII came much, much later with 100 megabit Ethernet and it's actually four 31.25 Mbps serial connections in parallel so that a 100Mbit controller chip only has to pump bits at 31.25 MHz (max) with most of it running at just under 16 Mhz which was fast for the day but still doable whereas a 125 MHz controller was out of the question. The 15 pin AUI connecter was just how everything was done at the start because the 10base-5 cable was usually in the ceiling to minimize overall length and because you couldn't just pull it through an office conduit like phone cable. When smaller networks of 10Base-2 started gaining traction you saw the backbones suddenly become departmental and venture into the cubicle or executive office and we got the pricey AUI thinNet transceivers shown in the video which were usually under the desk or in a 'wiring cubicle' which often held printers and common gear in a cubicle island. The next step was the cards with integrated PHYs, first for ThinNet and right after that you also saw pre-10Base-T AUI based transceivers for twisted pair but it was not yet the 10base-T of today and lacked the heartbeat signal. Prior to then your media choice was either coax big, or coax little. In 1990 10Base-T was ratified and suddenly actual 10Base-T to AUI adapters were common to reuse the old Ethernet cards with twisted pair. Finally in the early 90s twisted pair took off and coax became a memory but there were different PHYs for 10Base-T and 100Base-T and we started to get fiber so the makers couldn't just ditch the external PHYs just yet and it was implacticle is most cases (then) to put a dozen PHYs on a single die in a switch for example due to current consumption. By 1992 much was solved, but the ideal fab processes for a multi-PHY and for a switch IC remain different.
@@Peter_S_ Thanks for thar history. Makes me appreciate how really easy it is to implement hardware now in small devices for 100base-T.
Ah fond memories of Arcnet. I had upgraded my store from a DOS based network to a Netware 2.2 network in 1988 and with that came Ethernet. After about 2 months of one cable getting sliced all the time at one register I ripped out the Ethernet (Thin-net 10) and put in Arcnet. Took a bit of searching to find a Micro Channel Arcnet adapter for the IBM 50Z server but I found one. Arcnet was much more robust and could work when a workstation cable was cut or unplugged. Running it was easy and the speed was never an issue. Ah the memories.... now I manage a very large 100g network with thousands of hosts, both IP4 and IPv6, sigh.
One space after period is enough. This isn't the 1500s, we have proportional typefaces now. Two is already stretching it. Yours is just taking the piss.
This takes me back, Arcnet, Novell Netware, CSMA/CD. How much stuff you forget.....
Arcnet is still used in some industrial niche applications. (Mostly large machines and sites where changing to something newer would be cost prohibitive.) I took part in the design of new arcnet hardware for a customer just a couple of years ago. That was a fun project, having relatively recent hardware features like PCIe and USB3 right next to your ISA bus and arcnet hybride modules. ^^
It's impressive how much old hardware is around, simply because that is cheaper than buying new industrial or medical equipment.
Look up that Intel Quark developer board.
A P54C Pentium running at 400 MHz, PCIe, USB2, DDR3, Wifi, it even has 16 kiB embedded SRAM
@@HappyBeezerStudios Sadly I had never an excuse to play with an Quark, but they sound interesting! Yeah, there are still products I know of manufactured to this day that use 68k (the original DIL one, not newer types) and embedded 386, often obtained through more or less sketchy parts brokers. Simply because the product has some certification that is expensive to obtain. ^^ I am not generally opposed to that as long as the parts are not salvaged ones, but I question if that still hits the spirit of those reliability oriented qualification processes...
I was the Eastern US Arcnet specilist for Datapoint. I used to carry around a Compaq "luggable" with a special Arcnet card called a "Sniffer". (thank you Dr. Harry Saul). It had a special modification that would allow me to see all data from all packets. As a specialist, I would test customer installed network. Arcnet was SO forgiving (wiring wize)... One long weekend, I forgot to bring home some coax to connect a couple machines at home. I ended up just using some lamp (zip) cord... Stripped it back to wrap around the outside of the BNC connector and then sharpend the other wire and stuck it in the center core.. It worked for what I was doing. Great memories with Gordon Peterson... BTW, a little known fact, the internal code name for ArcNet while under development was "Internet".... Crazy times.... Great presentation and thank you !
I did not know about the internal code name.
Did you know any specifics about the proprietary black dipped PCBs on every arcnet card?
Quick story: in the early 90s we used to have mini-LAN parties and play network doom. Someone managed to get a handful of what we thought were ethernet cards. Couldnt find any DOS drivers so ended up putting out feelers on usenet (from memory). In the end, someone identified the cards as arcnet cards - that person was the inventor of arcnet, John Murphy :)
We never did get them working :D
One small detail.... Coaxial Ethernet was only 2 Mbit for it's standard speed. Ethernet didn't get 10 Mbit until it switched to twisted pair cabling. But you have told a great story and told it well.
And you hit the nail on the head re speed vs collissions.... When I worked in IT and supported networks the collision light on an ordinary Ethernet utp hub was ALWAYS on.... Like the collision light didn't blink... It was just on .....
I recently moved and while going through some old boxes I came across some spreadsheets where we recorded addresses for each card on a few networks I was responsible for back in the day. Ah, the memories! Passive hubs were evil - we eventually just changed them all out with active hubs; was far cheaper in the long run to just have active hubs everywhere.
And I vividly remember trying to explain the collision thing to people - you are absolutely right. In the end "big number" won out because it was simpler to understand. It was so frustrating back in the day.
Also on collisions the best analogy I used was CB Radio - since they were still pretty popular in the 80's. Two people talking at once = garbage; that's a collision.
When I was first introduced to Ethernet in a Comp Sci lecture in 1979, as I recall, the lecturer had to admit that engineers were already prejudiced against the idea of such a lackadaisical approach to bandwidth management. I mean, collisions, for goshsakes! But he also said that real-world experience was already proving, to the surprise of many, that these collisions mattered a lot less than you would expect in practice.
About a decade later, I remember a colleague saying that Sun Microsystems (great networking champions, as you might recall) had done a test involving two pairs of workstations on the same Ethernet wire, the members of each pair exchanging packets with its partner at maximum speed, and overall getting to 98% total utilization, even after the collisions between them.
So don’t try to say that “users didn”t understand collisions”, when their own experts kept pointing out the issue at every opportunity. Ethernet won out because it worked.
@@lawrencedoliveiro9104 I don't know about Sun, but DEC did such a report in 1988: "WRL Research Report 88/4: Measured Capacity of an Ethernet: Myths and Reality," by Boggs, Mogul and Kent. And yes, for many cases (though not all) Ethernet is quite efficient even with a lot of hosts at high offered load (i.e., all transmitting data). It ranges from about 70% for very short (64 byte packets) up to somewhere in the 90s, IIRC.
I worked with arcnet. Gosh forbid you didn’t have an accurate list of all of the address dip switches for every computer. It could take forever to figure out who had the same address on your network.
As I was watching it I was thinking "address conflicts must have been a nightmare to work out if you didn't have a list to work from".
@@Firthy2002 You didn't want to be the keeper of the spreadsheet, but if you didn't and other people weren't as diligent your life would be a living hell - so I was always the keeper of the blasted spreadsheet. But at least the network was stable!
Also cards with the switches out the back were evil - we only bought cards where the switches were internal for reasons others have talked about elsewhere: those pesky users! Networks are always easier if you don't have pesky users running around mucking things up.
In 1986 the first LAN I installed was a nondedicated Novell 86B system using a single passive ARCnet hub. We were initially going to use GNet but we switched for a reason I don’t remember. The main purpose of the system was as a file server to share Microsoft Word (for DOS) files between the four computers.
Future networks used ARCnet into the 90’s. Most of my customers were accountants and law offices. The star topology worked well with excellent response times. It wasn’t until the Ethernet switch arrived that I started believing in and pushing clients to 10BaseT.
The convincer was a school district with a 10BaseT network running Windows 3.1 (might have been Workgroups) on diskless workstations booting from the network. All the classrooms involved tended to power up at the same time and it was a severe strain on the network. The server handled it just fine because Novell did their caching very well.
I added a shiny 3Com 24 port switch to replace a hub for one of the rooms and it was an instant success. We never looked back.
"They also have a side hustle in producing PCBs." Whatever the sponsorship deal is, they're getting a bargain - this was one of the rare advertisements that's truly entertaining, literally LOL. Now back to watching the video....
Good watch, i remember the surge of network cards in the late 90s going from coax to 10 base-T with hubs to 100mbit switched. It seemed to happen real fast even at the time.
I remember many many years ago setting up an Arcnet system where I worked. I still have the crimp tool for the rg58 cable & bnc plugs. Oh the memories lol.
Same here but I was able to use the crimp tool later for antenna coax so it was not a total waste of money.
@@pctrashtalk2069 Oh yes, I've used the tool for other stuff, but remember at the time the tool costing a fortune (Around £100 from memory) and it the Mid 80's that was a lot of money lol.
Didn’t Arcnet use 92 ohm RG62? With the tiny plastic spacer to keep the wire centered?
@@mikebroich1487 To be honest, this was back in the middle to late 80's (nearly 40 years ago). You are probably right, but then I can't remember what I did last week lol.
One of my first clients in the early 80’s as a field engineer for a custom semiconductor company was Datapoint in San Antonio TX. I still have nightmares about trying to get someone there to commit to a custom design for their Arcnet card. No other high tech in SA at that time, so it was a long trip from Dallas to wait all day for many meetings that never happened. That company killed itself, sorry to say.
Thats a shame for datapoint, if they could have cut the price of their cards back then they could have killed ethernets ablitiy to catch up on price.
After the venture capitalist Asher Edelman purchased Datapoint he basically stripped it for profit, selling off any worthwhile technology and Assets then sold the husk to Datapont UK in June of 2000. In the UK Datapoint had specialized in Telephony systems alongside Datapoint products and they re-purposed the brand to service only this market. Moving from the old book repository building they had occupied for two decades in Neasden North London, the new Datapoint UK headquartered itself in Brentford, England and had 5,000 client sites in 41 countries worldwide. Unfortunately Datapoint (U.K.) Limited was dissolved on 14 November 2013 so the dream ended there.
I worked for them in the 80's in Neasden and they were a wonderful company with a real sense of family, with many staff keeping in regular touch today. Our team was still meeting once a year right up until the Covid pandemic restrictions, which speak volumes, happy times, sadly missed.
In the 80s, Byte Magazine columnist Jerry Pournelle became interested in networking and ARCnet seemed the way to go at the time. So the cabling was run in the walls of a portion of his house, known to his readers as Chaos Manor. By the time I became a friend of the family, these mysterious coax jacks were long unused and it was years later that it was mentioned that these were the remnant of the first network in the house. I'd heard of ARCnet but Ethernet was completely dominant by the time I got interested in networking, so I didn't recognize what it was.
I cant tell you how many of his books I just loved. Every few years I find my self reading "The mote in gods eye", and "The gripping hand" again.
Those Chaos Manor articles were so much fun. I remember his struggles getting Niven (he never called him Larry) to use a word processor.
@@strayling1 somewhere in one of Larry's books he said about computers that he just used whatever Jerry built for him.
@@greggv8 That was true up until the point in the 80s when it came down to just having a reasonably recent PC. I built a few of those for Larry in the 90s and a bit later, until he lost interest in gaming and a good off the shelf Dell or HP made more sense for fulfilling his needs.
When Jerry was first trying to get Larry onboard with word processing it was a remarkable thing to be able to say you had a computer at home.
I still use arcnet today. A line of fire alarm panels made by Honeywell use arcnet for communication locally between the boards and long distance between the panels (up to 4k feet if I remember right) on 18awg twisted pair. It works, most of the time :D
My first home network was arcnet implemented with Lantasic. Ahh, both a simpler and more complicated time.
I remember those CentreCOMM MAUs from my first IT job, working for a university in the 90s. As they fell out of use they began to by used as doorstops for self-closing comm closet doors when we needed to be in and out of them often while doing work. I once found a stylish MAU under the data center floor when cleaning up the years of cables left when equipment was removed. I found an end of a length of coax and gave it a gentle pull, it eventually stopping because it was clearly attached to something on the other end. Upon following it to the other end I found a MAU in beige and light brown with a quarter round profile.
"Invent the future then don't tell anyone about it because it might hurt the copier business." Best summary of Xerox PARC ever.
I had heard of ARCNet long ago, but never really knew anything about it other than it existing. And especially never heard of the Victor Moore/8008 (guessing that was supposed to be the 8008 rather than the 8088)/Intel bit. Incredibly fascinating and educational video, downright eye-opening, and as always your visuals are fantastic. Thanks!
Vic Poor was the name…
Yes, should have been 8008. And Victor Poor was a self-taught tech genius, and vice-president / tech guru at Datapoint. He, and a young fellow / ham-radio guy named Harry Pyle created the 8008 architecture over Harry's Thanksgiving college break. Both have oral histories online, recorded by the Computer History Museum.
I remember mentions of ArcNet in the late 80s and early 90s relating to networking on the Amiga. Although I've never knowingly encountered an ArcNet network myself (my high school was on an Ethernet network in '95). I discovered the lack of ArcNet-Ethernet routing hardware myself not too long ago when I was looking into how someone might get an Amiga onto a modern network.
I was trying to see if cisco had done an arcnet card for any of there routers but could not fine one. I do have an arcnet card for an amiga 2000, but I could not find where I've put it.
Netware had a built in software router - that's how most arcnet to ethernet network connecting happened back in the day.
Fascinating. I had never heard of Arcnet, nor knew that early Ethernet cards couldn't connect directly to the network. Thank you!
You're welcome.
I remember experimenting with some old ARCNet gear in the early 90s. My instructor once said "you can run ARCNet on barbed wire!"
Well, I never had barbed wire, but I did manage to run it on coat hangars and a piece of broken Ethernet cable.
coax ethernet was fairly popular in some uses to late 90s. not having to set station addresses and not have switch ports for everyone was pretty useful at higher headcount scene/lan parties. sparks ahoy
It was a real pain when something was not working because none of it worked.
When runs were longer I ran into cases where station A could talk to B, and B could talk to C at the far end of the coax, but A couldn't talk to C. Some cards handled the long runs better than others. I watched it happen with an oscilloscope... Sure did like the simple cable runs and termination though :)
At my last employer, we had a Technicians’ Department with some folks who knew their network hardware stuff. (My specialty was in software things Apple- and DEC-related, including configuring AppleTalk protocols.) One of the gadgets they had was a Time-Domain Reflectometer (aka “TDR”) which was handy for finding discontinuities in network cables.
Another excellent video. Maybe use the spinning blue network background a little less though. It's rather dizzying when your trying to look at a still image with a spinning background. Keep up the awesome videos.
In the late 1980s while I was an undergraduate at the University of New Mexico, I worked for a small company that eventually deployed a Novell Netware network using Arcnet to connect a growing number of 8088 PCs to our 80386 file server to run dBase III+ applications. It worked well for what we needed, allowing any of the workstations to access the database on the server and run independently.
As a Computer Science major, I mainly had been introduced to OSI and IP networking and to a lesser extent Ethernet prior to the company installing Arcnet, but I can see the advantages, especially for a small start-up in that era.
dBase III+ now there is a blast from the past, I remember getting a NLM (maybe for a later version of dBase) and getting that installed and running that allowed for much better locking. It stopped our occasional bits of database corruption, and improved performance for us as well. Its a shame the way dBase lots most its market to Access.
OSI networking ... you poor kid ...
Arcnet isn't dead. It's still used to connect HVAC systems to building management system controllers.
The stuff CTC was doing seems amazing at the time. .. distributed storage and processing, integrated data and voice .. with micros .. in the late 70s and early 80s. We didn't get that where I work until 1997
My first adventures with a home LAN were with cheapo Realtek NE2000 compatible cards and coax cabling to the next floor so that we could do some LAN gaming. Quickly upgraded to a 10/100 hub, though, because the whole coax cable/T-piece/termination thing was very flakey, probably because I could only afford the cheap noname stuff. Upgrading later again to a 3com officeconnect switch and 3com 905TX cards was a revelation!
These videos are fascinating, I started my networking journey in the late-90s, and everything was Ethernet in the places I worked. So learning about the history of networking and different standards is immensely fascinating to watch.
Less than 3 minutes in and you have already blown my mind with the x86 connection
I remember one factory with arcnet that had a wall that was over a foot of concret and a steamline throughit and we connected the arcnet line on each side and the damn thing worked great. That and it was the only way to get a network connection to that part of the factory much easier than trying to get a hole drilled.
So, just to check if I understand correctly, you ran the cable through a pipe that carried steam through that wall? :-)
@@SatumangoTheGreat no we attached the cable to the pipe.
@@DXMage Ah ok. Thank you for the explanation.
@@SatumangoTheGreat yeah it was the only way short of running a cable hundreds of feet and well going through massive amounts of work my lead asked me for ideas and I said 15 muntes or so for us to test it and if it works then we save a HUGE headache customer saves a hella ton of money we all go home for the day if it doesn't we are in for at least three days work at least and even then it is going to be a major pain in the butt. We tested the signal was well within spec on the other side and it worked flawless. Definitly not a recommended way to do it but it worked =) OH and drilling a hole wasn't an option as the customer made that VERY clear.
That issue about not having to manage station addresses is a big win for businesses, especially those that didn't have dedicated IT management teams. Ethernet is simple; you just plug in a new device and, generally, it just works. Plug and play is the thing. Of course when it became just twisted pair and hubs then switches that made it all very easy.
It's also a lesson that IBM didn't learn as SNA was the opposite of plug and play. It was very secure of course, but it required armies of administrators and IT department bureaucracy to set up your printer. Even in large corporations, there were semi-autonomous units that simply didn't appreciate all the delays in getting the IT department to get their new printer working on the network.
You're right not having to manage station numbers was a big win, once ethernet moved to utp the problems with its cabling randomly breaking and you having to follow the cable around the buidling went away, which then made ethernet the less hassle option.
For a long time, SNA wasn’t even a “network” as we would understand it today--namely, the whole peer-to-peer element was completely missing until a later version of SNA came along sometime in the 1980s to play catch-up to TCP/IP, AppleTalk and all the rest. The new version was identified by acronyms like “LU 6.2”, “PU 2.1” and “APPN” (“Application Program-to-Program Networking”, if I recall).
@@lawrencedoliveiro9104 You have to love IBM and their acronyms/abbreviations. However, I have a soft spot for the mainframe architecture as I started out programming using IBM 370 assembler code writing operating system code (for a non-IBM OS as it happens - a long story).
As far as SNA goes, and other things IBM, I recall a lot of what might even be called denialism on the part of the IBM technical support community when open systems came along.
So many of them had gone through the way that IBM trained system programmers, DBAs and the like that they'd almost been indoctrinated into thinking that the way IBM did things was the only way. Their world was being turned upside down by what looked like chaos. Even when they worked on flawed things like CICS, CKD architecture disks, and the wonderful complexities of VSAM.
I remember Arcnet! Ahh good times supporting a site with an 'eclectic' mix of kit. The organisation (a very large, rich one) bought stuff on Ebay to add stations or for spares. Every so often the couriers would bring in job lots of assorted PCBS, dumb terminals, chassis etc in boxes and it was a case of "Make something work".
ArcNet lived on as the basis for a number of industrial control system network protocols - the reliable timing and intrinsic confirmations made it very suitable for distributed control systems.
It would be interesting to hear about the evolution technologies such as CAN Bus and how they fitted into the evolution of networking.
Belt-way scale totalizers used ARCnet less than a decade ago. We still have some running in the field, and I have NIB spares...
The content in this video is excellent. I worked with Arcnet for many years. I still have computers that an use it.
Nice. I vaguely remember going through some arcnet-training at Xerox 23 years ago, together with Novell netware and some other things I never had any use for.
I got my Certified Novell Administrator certificate for Netware 3.x, just in time for the debut of Netware 4.x. I just bought the big Netware 3.12 book, read it cover to cover, then went to the local Sylvan testing center and passed the test. Wasn't long after that Sylvan and Prometric merged to become Sylvan-Prometric and as most monopolies do, they massively raised the pricing for their Netware and other certification tests. I *was* going to buy the big book for Netware 4.x and get that CNA cert but when I saw the price for the test I said hell with that. It would have been a waste because Netware ended up being a dead end.
I remember faffing about with hardware in the 286 days and finding what I thought were thin-net cards as they had a coax connector and no AUI port. I had always assumed the dip switches were for setting IRQs/Ports. But after seeing this, I guess they were ARCNET
My first "real" job was for my hometown school district. Novell network and tons of arcnet. Even as other portions switched over to 10base2 or 10baseT longer runs between buildings (grounding issues aside) were only possible with arcnet. That worked until budget was finally available to make those runs fiber. :-)
I remember getting our first fibre link, it felt so big budget at the time. We had to get due to grounding issues between different parts of the building, there was a 50v potential difference, which was sufficient to kill equipment, and hurt.
@@RetroBytesUK a friend used to do internal IT support for Albertson's (one of the largest grocery store companies in the USA). At one location they had a gas station in the parking lot. The system in that small building connected to the server in the main store using an Ethernet cable run through conduit. Periodically they'd get a lightning strike in the center of the parking lot, right on the network cable. Various computer equipment at one or both ends would get fried. (IIRC they'd also have to pull new cable through the conduit.) After a few times of that the company upgraded the link to fiber optic and never had another lightning strike in the parking lot.
As a bod who has worked in loads of warehouses they never change anything unless its 100% dead i was still using a DOS pc to print lables in 2016.
Mine went down 3 years ago. I felt sad. 😢
One of the great success stories that contributed to Arcnet demise, was 3Com's & Ungerman-Bass fast Ethernet switches were were very cost effective.
There where very cost effective for the time, being able to link the back plane of 3 of them together really increase how usful they where too.
Arcnet was my career for a few years and it is installed in some very interesting and quite secret facilities.
Such a shame they did not do a much faster version but this is why having one corporation with a monopoly is a very bad idea.
Didn't realize that the phy interface to the mac was external to the card on early implementations. Helps explain why mii and rmii are part of the 802.x standards.
I shortened the bit about how the transceiver started to get included on the card, and just moved to the NE2000 as that was the first card with it included most places first used. There where other cards that did it first including the NE1000, but the NE2000 (which was basically the NE1000 with DMA) was the one that sold in volume where most got to encounter it. Someone in the comment has given a much more detailed history of that bit than I did, as I was trying to not make the video too much about ethernet.
@@RetroBytesUK Ah, you've sucked me into looking at this ancient history. Looks like that wasn't the mii interface. Seems like it was probably the inspiration for it. Motivation to separate the analog parts of the phy from the purely digital functions of the mac.
My company was heavily invested in token ring back in the early 1990s! We ended up replacing it with Ethernet.
While my mid-90s high school had a network based mostly on Token Ring, the typing lab did use Arcnet.
Arcnet was my comms of choice for ROV comms - I could get 2000ft of communications to the sea bed in the North sea on a twisted pair working at a depth of generally less than 1000ft - then it was picked up by Tritech for it's stability (token passing) for their sonars and profilers etc. good engineering gets selected by good engineers, not accountants, they listen to sales lies and crap. However 10 Gbit and faster rates have rendered this moot on fibre optic systems. these are reaching more than 11km the bottom of then Marianas trench (approx. 10,880 metres). Standard Moog kit 907 types guarantee 10Km reliable comms now.
This was genuinely an interesting history lesson! I've been an engineer for 20 years, never heard of this standard before, Nice!
I serviced some businesses in the late 90s and early 2000s with these lol, I learned about it while in the field as a tech for a large warranty service company, don;t really miss that.
I remember seeing Novelle Netware installed on some school computers back in the early 2000s. Time really flew by.
I once worked at a company where we have Novell servers with both Arcnet and Ethernet cards in them because the back-of-house was all on Arcnet where as the front was Ethernet (10-Base-T coax), so the file and print servers (back when those were separate things) had to be able to talk to both networks. We also had dumb terminals for a Unix system that where connected by serial cables to concentrators where said Unix system was connected to the Ethernet (mostly so it could use the Novell print server). There was also a small Token Ring setup in one of the test labs, so we had it all. Ah, fun times.
Another idea for a video would be discussing now obsolete network server software, such as novell and banyan vines, etc.
An early 100Mbps token ring option for rich people was FDDI ("fiber distributed data interface" defined by ANSI X3T9.5 in 1984)
FDDI was a very interesting technology, I only got to see it deployed twice. Once on a site that used token ring for the clients and an FDDI ring for the servers. The other was a cable company that used it to link their street cabinates prior to using docsis, it was the backhawl for all their telephony services.
Nice one! Thanks. Learnet quite a bit about ArcNet. One thing, though: you mentioned Token Ring having 8 or 16 MBit. But: Token Ring had (on the PC) typically 4 or 16 MBit/s.
There as been other TokenRing implementations, too, e.g. 12 MBit with Apollo Domain workstations. But the beforementioned 8 MBit/s Token Ring is either quite an oddity or not existing :)
OMG. Another Domain-iac on a TH-cam. Too bad I can only give one thumbs up.
We had a giant Apollo token ring network with the office split into multiple sub-rings that could be switched out/in in the wiring closet when we needed to add connectors to the sub rings
@@ppokorny99 I never experienced Apollo Domains in their natural habitate and time context (it’s not my fault! I’m born too late!) but only years later as nice workstations. The 2000 and 2500 now serve in our Regional Retro Data Centre, alongside RS/6000, Acorn Archimedes, Apple III, Commodore 8032, Atari Mega ST and many others. Coming back to your point: I still need a way to route packets between Apollo TR 12 MBit and the other network systems we operate (FDDI, 4/16 Token Ring, 10BASE-T1L, 100VG-AnyLAN, IBM BROADBAND, LocalTalk, FlashTalk …)
What was the main purpose of your Apollos?
Came here to say this, comments did not disappoint. I have put in brand new 4/16 mbit real IBM microchannel cards back in the day, so quite familiar...
I'm always into videos that start with ripping Intel a new one... :) Also, I remember, at the mid-to-late-90s, I was still running into ARCNet networks as a computer-wiz-kid, but there was very little info available (to me), and I already had some Ethernet knowledge, so the troubleshooting was usually just to rip the thing out, and replace with the by-then dirt cheap Ethernet. But I was always intrigued by it, like some mystical stuff that was built by "ancients". So today, after all these years, I learned something. Thanks!
My first home network was Arcnet.
Wow... Mine was ipx. This is the first time I ever felt like I was late to the party.
You have never lived unless you got to hear a 12 port ArcNet hub do the tap dance of death while everyone is screaming 'their computers are broken'.
I've heard tell, never got to experience all the relays got nuts my self however.
That video brought back many memories. I worked for a major financial institution in Canada who inbitially had a large Datapoint deployment; both with true Datapoint hardware and third-party servers from Performance Technology,. Then a Banyan / VINES network was added in to the mix, all running over ARCNET. The joys of managing the network address on each PC and of ensuring we didn't exceed the max number of stations on a single ARCNET network. Some areas did start to use ethernet, and of course there was also Token Ring.
I haven't heard those words in 40 years. Banyan / VINES. connecting remote offices over modems on the vines network. Magical times.
The problem with the throughput topic is that while the theoretical “worst case” throughput of 2.5Mb ARCNet is higher than 10Mb Ethernet, the conditions under which that would happen have NEVER been demonstrated in the real-world. Collisions are uncommon in any demand network below 50% utilization, which means that until you see average demand on a “thin-net” exceeding 5Mb/s for an extended time-period, you’re not going to see any degradation. At that point, you’re already at 2x the hypothetical performance of ARCNet, and you don’t reach saturation until 70%-90% in real-world situations, so the real-world performance of 10Mb Ethernet is between 2x and 3.5x the maximum throughput of ARCNet.
IBM’s Token Ring suffers from the same limitations, even in 16Mb version. In 1988/1989, I did real-world testing of 10Mb Ethernet (thin-net) vs 16Mb Token Ring, complete with MAUs, on 286/386 workstations and a 386 based server. In every single test, single station load and many station load, each station attempting to transfer as much data as possible, the 10Mb Ethernet outperformed the 16Mb Token ring in both single station and composite multi-station throughput. And Ethernet was significantly cheaper too.
100Mb Ethernet then made Token Ring completely irrelevant.
Actually, the (incorrect) figure that most people pull out of the theoretical studies and modelling studies is that Ethernet tops out at 37% offered load, which at 3.7 Mbps is already 50% higher than ARCNet's maximum speed of 2.5 Mbps.
But that 37% figure doesn't actually bear much, if any, relation to how real Ethernet works. If you want all the gory details, the the 1984 paper Boggs, Mogul and Kent, "WRL Research Report 88/4: Measured Capacity of an Ethernet: Myths and Reality," is a good read. Note that the saturation at 70% figure they measured in their study was with 64 byte packets, which is unrealistically low for almost any situation.
Great explanation of ARC Net. I still have here a few 3Comm ARRnet/Ethernet ISA cards. Tucked away in my archive of "old DOS" stuff is a version of Netware Lite that you have to load the protocol drivers in the correct order to get it to work. Was fun.
My first job in computing was working for mintex the brake and clutch manufacturer. we had sales depots around the UK each was equipped with datapoint computers recording sales made during the day and generating invoices. My task was to use dial up technology to obtain the sales records, run a programm to update stock records which then went on to generate restocking levels and factory demand. Most of the 23 depots had 3 floppy drives on the computers where the head office where I was based had Winchester hard drives of around 10 mb to store the information. Backups, report generation etc. When the system went wrong usually in end of day updates or data communications I was the one who remotely working in hexadécimal to write end of records and close files. 5 years experience lead me into teaching and I never looked back. Thanks datapoint and mintex and of course dos. Happy days
I worked with both Datapoint and Netware. 3Com introduced ethernet cards to offload the network processing to the card such as 3C503 and the server 3C505.
Yup, the company I worked for in the late 80s pulled out our Arcnet for 10Mb BASE-T around 1989 when we brought in a new SCO/Oracle system running on a Tricord and never looked back. NetWare286, modem banks, Telebit dial-up routers, ODBC and Lotus 123... Fun times.
I wired one school with the ArcNet+ network in 1992. As college had taught only very little outside Ethernet, I was first a bit puzzled as to what it was. However, crimping coax cables was a breeze after a handful of failures.
Lived through this running a computer network for my university from 88 - 92. With our 8086 and 80286 labs, they were all ARCnet. By the time the 386 and 486 machines were coming around everything was Ethernet. Novell made a “MPR” multi-protocol-router that would bridge our ARCnet to Ethernet (IPX -> TCP was our use case).
Price was the biggest factor driving the convergence to Ethernet. The other driving factor was the emerging internet and TCP/IP. ARCNet’s limited TCP/IP stack support in most PC OS’s was a huge drawback. Then you add Windows NT server and Linux starting on the scene and things get really grim for Novell and by extension ARCnet.
You didn't mention that the big reason there weren't lots of companies making arcnet cards was because Datapoint required every manufacturer to buy those proprietary black coated PCB modules from them. That's the rectangular piece close to the BNC at 12:00 As a single source supplier for a must have component, Datapoint could pretty much set the price that arcnet cards would sell for.
The 1989 introduction of the Ethernet switch by Kalpana rang the doom bell for most other network technologies. Switches could be swapped in place of hubs and instantly boost aggregate throughput on a highly loaded Ethernet network, while requiring no reconfiguration at any of the connected computers or other devices.
I had one large pre-installed arcnet system that was upgraded to 10 mbps Ethernet. To bad it was before twisted pair was used. So we used coax 58ohm? (to many years ago.) Also had one token ring network existing. Added a 2nd card in the Novell server to expand with Ethernet. But, that was the life when I started into computers. I found out later on Arcnet was already refereed to as pcnet. Then I found a few other cool things about it. Like radio shack used it for a trs-80 network.
The coax you are taking about is 50ohms impedance. Always a problem if someone pinched the terminator at the end of a run.
@@gorillaau Ah, to many years ago. My fav was them unplugging x amount of computers and having it go down. So happy when Twisted Pair and hubs came around.
The 8086 is pretty district architecture from the 8080. Although you could argue that it's a combination of ideas from the 8080 and the 6502/6800 processors, so not entirely unique.
Also despite what you may have heard, the 8086 is not source compatible with the 8080. This is pretty obvious when you think about it as the 8080 has a bunch of instructions that the 8086 doesn't and they have different register names. I believe this misconception comes from the existence of programs to convert code from one to the other. Over time "You can convert between then" becomes "They are compatible".
Datapoint, now there's a blast from the past. I started my IT career back in the '80s working as a programmer for a UK software company writing in Databus - those were fun times.
I encountered Arcnet in an embedded application in about 2002 I think. At that time it had two incompatible physical implementations, one using what I think was the original AC transformer coupled implementation and a second one using RS485 trancievers. Unfortunately this application used the RS485 variant and the link shared a cable with a 56V DC supply, so the trancievers died frequently.
I was asked to repair the boards so they could go back into service. My first "fix" involved fitting Maxim "fault protected" 485 trancievers which could withstand the full 56V they'd see in a worst-case misconnection.
So I thought ... then I'm presented with units that will communicate with the host PC directly, but not over the long cable. Turns out the terminating resistor has ceased to exist. So now I fit a PTC thermistor with a cold resistance near enough to 100 ohms.
Meanwhile a newer generation of units use the AC coupled version and later a variant with a DC-blocking capacitor so a misconnection can't burn out the transformer.
As an experiment I tried adapting their two-pair cable to 100baseTX and I could run 100baseTX over it and with a bastard POE implementation could have run the DC over the same cable with half the voltage drop of the original system.
I remember having to use a vampire tap to connect a computer to a thicknet cable. The cable was marked to show you where you could put a tap. When you were having problems with thinnet, it was usually because you were missing a terminating resistor at the end of a line. People would move desks and computers and weren't paying attention to the cables. This was in the early to mid 90's.
I was always jealous of the NE2000 cards because my company ran Banyan Vines servers. Now THERE is a topic for a future video. Vines required special cards at first. It wasn't until the 3Com 3C503 cards appeared that we got a cheaper option.
I never got to see vines in its special card phase only after you could use etherent. Some times I feel like the only person who remebers it, as I get very blank looks when ever I bring it up among old networking types.
We used ethernet. The cards just had to be Vines compatible. Looks like they added NE2000 and TCIP/IP support with the last versions.We had moved to NT by then. For a brief period, Vines was the only NOS that could scale REALLY big.
Banyan was also one of the first LAN companies to introduce their own directory server (think X.500, LDAP, Active Directory etc).
Microsoft hired Jim Allchin away from Banyan to bring StreetTalk to LAN Manager. A couple of years later we got Active Directory. So in some ways it is still around.
I maintain an old novell network running on dos... they refused to upgrade decades ago, so it's job security, I guess.
Year 2000 (Y2K) retired the Netware 286 systems I supported. There were a few Netware 4 systems that had to have hardware die to get them to their heavenly reward.
@@richshealer3755 ebay, someone sent her a working motherboard with an ISA slot. and... it was alive again. **sigh** someday it'll all croak off.
It was my job at CompuPro, in the 80's, to make the Arcnet adapters and software for both Pc's and the S-100 machines. The most difficult thing was to make the state machines on each node work, when the protocol was not a symmetrical dialog process, as is the case with Ethernet. That meant that the token bus approach was inherently less robust against signal dropouts or packet collisions.
So, the possible event points were: reconfigure after reset, waiting for token, getting token, releasing token, rinse repeat. Or, get token, send packet, wait for ack to sender, get ack from recipient, wait for packet reply from recipient, send back of reply from sender to recipient, rinse repeat. At each of these points a reconfig event might arrive asynchronously, requiring resetting the sending or receiving packet counters kept in each node, except for the times when it doesn't. And theassymetry occurs at the time interval between the first ask and the packet reply. There is where the most danger lies for losing sync of the packet counters.
ARCNET also supported a loop through cabling topology using BNC Tee connectors just like Ethernet on Thinnet. Exactly the same except that ARCNET used RG62 coax cable at 93ohms and Thinnet used RG58 coax at 50ohms. Hubs were not required for Arcnet but they did did make cabling easier and the network reliability was also higher as the loop topology is easily broken at any attached computer taking the whole network down. Note too that you could also have both hubs and looped segments with ARCNET. I was a big user of ARCNET in the 1980s and 90s for industrial computer automation networks. I even designed and fabricated custom ARCNET embedded hardware. I miss those days!
In 90's I participated in a project that incorporated Arcnet into an embedded network. Included was also a proprietary ISA-card for connecting a PC into the system. Those were the days.
At the time, Intel never claimed that they had invented the microprocessor. In fact, they didn't even attempt to patent it, but said instead that there's nothing patentable in a microprocessor, since for something to be patentable, it needs to have a non-obvious original idea and the idea of a microprocessor was obvious to anyone working in computing at the time.
I love this channel. I vaguely remember my mom messing with some Arcnet stuff when I was a kid. And I don't mean she was in IT, just she did work on machines hooked to an Arcnet network. I remember her working with and around all kinds of stuff that is rare now or was rare even back then. Or was once popular and died off because something else took its place.
Arcnet was the initial choice for Tandy's first networking forays (TRS-80 Model 2 and 16). When I worked there, it was so slow (even for the times) that internally we called it ArcNOT and Bark-net
What a deep subject I never knew I wanted to care about 😂 Any thoughts on making a video covering Unix in general? What it is/why it’s important/how it came to dominate early on/etc?
My best friend in higschool and I ran a two-node Renegade BBS on a 486SX-25 and a 386DX-33 (I believe), and networked the two nodes via Arcnet. Didn't have the right connectors, so we used coat hangers for a while...
I worked for a company that was supposed to be focused on error proofing everything. That’s what they said. What they did was connect both Ethernet and arcnet to the same operator station. The cables looked identical, but had different impedances. The terminators had the resistance stamped on them, but they were hard to read, so they were often installed on the wrong network. Bottom line, the systems never worked out of the box.
Thank you for a really interesting and in-depth look into a technology I knew little about.
Fascinating and well told, thanks for sharing dude!
Glad you enjoyed it!
Oh damn, finally you've done a subject that I've (thankfully) been too young to deal with
Great video. And one thing that wasn't mentioned as a nail in the coffin is the introduction of the Ethernet switch. It mitigated the impact of collisions in many scenarios.
I'd like to see a RetroBytes video on PCMCIA and CardBus, and how it worked in DOS + Windows 3.x. Getting it to work in DOS required a lot of drivers. First there was the Card and Socket Services driver specific to the PCMCIA chip. Then for the cards there could be either Point Drivers or Class Drivers. A Point Driver worked with one specific card, for example a SCSI controller, sound card, or other device that had unique properties. A Class Driver worked with many devices of the same type, like battery backed DRAM cards or NE2000 compatible Ethernet cards.
The real trick was getting the C&SS driver plus Point and Class drivers for more than two cards all crammed in simultaneously (no dynamic loading and unloading of drivers) while also having all the DOS drivers for everything else in the laptop AND still having enough free RAM to run Windows or a DOS program. But with it working one could hot swap the cards that had drivers installed. If you needed more than 3 or 4 different cards needing Point drivers you had to setup a multiple choice DOS boot to select different sets of drivers.
I don't know how well Windows NT4 handled PCMCIA but I do know that Windows NT 3.5 didn't get along with it on at least one laptop. In the late 90's a friend got an old laptop and just for fun he installed NT 3.5 on it and tried for quite a while to get the PCMCIA slots to work. Nope, wasn't happening.
Windows 95 erased all those limitations. It needed no DOS help for PCMCIA, though without DOS drivers loaded DOS software couldn't use the PCMCIA slots. IIRC the laptop I had with Windows 95, I had 4 or 5 different PCMCIA cards I used in it without a problem. It also had easily swappable hard drives so I did one with a stripped down Win95 install with Basilisk II as the default shell so I had a Dell Powerbook that booted into Mac OS 8.1. Since the Mac emulator was the shell, shutting it down or restarting it would make Windows shut down or restart. It was mostly a "Ain't this neat?" deal because it was a 75 Mhz Pentium with passive matrix LCD.
Don't know which video was the first one that came into my feed a few days ago and had you taking me down memory lane but thanks. I got my first micro in 1983 when I was 14, idiots at school called my dad because I wrote a paper on computers and the staff thought I plagiarized it. It's funny watching history and documentary stuff and being like "I remember that" and "I was screwing around there". In a way it might be a bit like how some of the engineers and workers on the 1950s/60s rocket assembly teams felt. Even got to do some work on a mainframe before Y2K, the only people making a fuss about that was the press. Everyone in the field know it was just a quick cash grab because the pres wanted to turn it into something they could hype to up their ad rates.
This misses the whole effect of IP which was used on most ethernet but not ARCnet. Ethernet allows for use in a WAN while ARCnet is really LAN only. IP was really the driving force which is why ARCnet is only seen on factories. Thick wire and thin wire are also confused here. Interesting overall however.