Right now, I'm reading Cory Doctorow's 2024 book "the Internet Con - How to Seize the means of Computation" which has a number of relevant insights, IMNSHO - https://www.versobooks.com/products/3035-the-internet-con Interoperability turns out to be a big deal... What do you think?
Interoperability is a huge deal. If the big boys could eliminate it and lock us in they probably would. The problem for them is that there are too many people who would need to agree for it to work, and then they'd need to eliminate the competitors. As it is though google and meta have tied up ad spend and the like which is extremely unhealthy
I find Doctorow to be interesting and insightful but with enormous blind spots. For example he completely fails to grok the second amendment. But on the other hand enshittification is right on the mark
That's an interesting blind spot... Suppose if the people of the UK and Ireland were armed - would the government bring in huge numbers of migrants, facing a lethal backlash? Or pull off the "vaccine" tyranny - same case for Australia, btw - all three had put in extensive gun restrictions ten and twenty years ago And the only means of redress is protest which is easily ignored, there's no possibility of switching out governments - which Jefferson posited as not only a right but a duty: "but when a long train of abuses and usurpations, begun at a distinguished period, and pursuing invariably the same object evinces a design to subject them to arbitrary power, it is their right, it is their duty, to throw off such government, and to provide new guards for their future security." https://www.let.rug.nl/usa/documents/1776-1785/jeffersons-draft-of-the-declaration-of-independence.php and that's what the Second Amendment if all about, it was designed as a bulwark against tyranny in government... Not having this right increases the "switching cost" to something out of range... And not only corps, but governments also are subject to enshittification...
Loved reading this. Not a programmer, but when I went (back) to grad school, a workstudy job was running the LAN for my dept, thru which I met a kindly gent running everything in the Dook dungeons. My job was boring and MG suggested I type "rn" at the *nix prompt. What a world! When I say the list of ahem newsgroups could be printed on about 10pp, you'll know it wasn't the beginning of the world but I could see it right behind & talk/listen to people who had a hand in building it.
Among others, I read alt.folklore.computers daily for years; frankly everyone I read conversed easily in these topics (so I understand more than I can do), and your piece took me right back to those days. Disclosure, big fan of esr, not least for his ability to explain things and willingness to change his mind, e.g., https://x.com/esrtweet/status/1910809356381413593 led to an epiphany of my own. In the thread you cite, perhaps you'd agree he implicitly acknowledges general correctness of what I take to be at least one main point of your comments (responding to https://x.com/CDoombeard/status/1982168304379728078). Anyway, thanks for the read.
"in a large busy network of several dozen or more devices Ethernet became unusable"
I was at Xerox and then at 3Com. Both had very large numbers of peers on their Ethernet, were not "client-server" and "unusable" never happened. Please substantiate this. It's basically marketecture.
Novell COULD work on Ethernet but also on ARCNet, Corvus, or any other hardware people had. ARCNet was actually quite successful for a while.
Netware worked because it tended to break Ethernet into small segments that were routed. The same with TCP/IP once Cisco and Wellfleet released their first LAN routers. Small ethernet segments with almost all the traffic going to the router worked great. The catch was the router was more complex and expensive than a token ring bridge.
I don't know what the network at Xerox was like, but I do know that the Cambridge University implementation with lots of Xerox workstations turned out to be heavily server based as email and file storage was centralized. So yes the computers could talk peer to peer, but in practice 90% of the time they didn't. In fact once of the things Cambridge learned was that XServer sessions from one workstation to another broke things so much that they did ugly hacks like add a second NIC on a dedicated LAN for the workstations where people really needed to do that.
I expect 3com was the same. In fact IIRC 3com 3+ Open was a classic client-server system. Peer to peer sounds great but it turns out that except for chat apps most actual use cases are in fact client server.
The only thing I'd note about this was XServer (or X Windows) came much much later than the PARC Ethernet or 3+Open: late 80's or early 90's. In fact, calling a personal workstation a "server" seemed backwards to a lot of people, including me.
The times weren't that far apart. I know Cambridge had X in the late 1980s because among other things I had a project to make an Acorn Archimedes an XServer over a serial connection in 1988? (I think, maybe 89?) That was more or less the same time as I interned at Madge and reverse engineered a chunk of Novell's IPX protocol. That was the same time other Madge people were working with 3+Open and the Microsoft equivalent whose name escapes me.
I was doing contract work for Novell back in the day, writing up white papers and such for their product line. I went to Brainshare 2000, where there was a huge number of attendees. NetWare ruled the LANs.
Then NetWare 6 had its big release day on a particular Tuesday: 11 Sep 2001.
Bad omen, I guess. Then Microsoft came along and ate its lunch. Novell catered to the devs, but Microsoft wined and dined the suits.
It sounds like yet another example of the best tech not winning the day. Makes me sad.
Netware had issues from Netware 4 which took too long to get everything working properly with Win 95. That let Microsoft get the bugs out of NT Server and also let TCP/IP get a foot in the commercial door. Once MS put everything on IP, IPX was dead and that eventually killed the Novell model
I won't pretend to be a networking guru, but I've had to set up office networks in a few startups. One thing I learned early on was the technical differences between a "router", a "bridge", and a "hub".
Routers create a sort of network partition, and are smart enough to keep traffic inside their local network local, while forwarding traffic that needs to go out to the "outbound" network.
Hubs are just dumb repeaters. They often look like routers - and are a lot cheaper because they're dumber - but repeat traffic out to everything that's connected to them.
Bridges are basically bigger, more scalable, and more expensive routers that you need if you have multiple office networks that need to interact with each other.
Too many people would see a $15 hub and just fill their office network with them, and wonder why their office networking was so slow, especially when 10baseT was what most people used. A few times I pretended to be a networking guy just by figuring out where routers should be instead of hubs, buying a few $50 routers, deploying them, and suddenly the network is decently fast.
As for IBM worlds, the one thing that scaled better than IBM networks was IBM expense. A couple of startups I was at briefly chatted with IBM sales reps, got basically "the $2000 you are budgeting for your 50 device office network will be like $15K for an IBM network", and that was that.
Yes a few routers solved most Ethernet problems and IBM was never the choice for the budget conscious. That was why Token Ring failed IMO. These days the "you can buy better, but you can't pay more for it" flag has been handed off to Cisco
Well, I always like to see a non-expert guy solve a problem that Ph.D.'s and grad students struggle with and write whole textbooks on. Good for you.
IBM was used to charging ultra-premium prices, and counting on executives to say, "No one ever got fired for buying IBM." In the 80's that magic disappeared for good.
The Micro Channel (late 80's) was their attempt to kill the clones and regain control over computing. Amusingly enough, they paid 3Com (where I was) to develop an Ethernet adapter for it. So even then, they were realizing that the genie was out of the bottle.
Microchannel was yet another example of IBM trying to be sort of open standards but proprietary and screwing up as a result. MCA was objectively better than EISA but because they wouldn't license the bus to other PC makers it was destined to fail because IBM PS/2 servers were several times the price of almost as good Dells or Compaqs.
Needless to say a ton of the ideas in MCA were adopted by the subsequent PCI standard and IBM eventually ditched MCA for PCI.
Right now, I'm reading Cory Doctorow's 2024 book "the Internet Con - How to Seize the means of Computation" which has a number of relevant insights, IMNSHO - https://www.versobooks.com/products/3035-the-internet-con Interoperability turns out to be a big deal... What do you think?
Interoperability is a huge deal. If the big boys could eliminate it and lock us in they probably would. The problem for them is that there are too many people who would need to agree for it to work, and then they'd need to eliminate the competitors. As it is though google and meta have tied up ad spend and the like which is extremely unhealthy
I find Doctorow to be interesting and insightful but with enormous blind spots. For example he completely fails to grok the second amendment. But on the other hand enshittification is right on the mark
That's an interesting blind spot... Suppose if the people of the UK and Ireland were armed - would the government bring in huge numbers of migrants, facing a lethal backlash? Or pull off the "vaccine" tyranny - same case for Australia, btw - all three had put in extensive gun restrictions ten and twenty years ago And the only means of redress is protest which is easily ignored, there's no possibility of switching out governments - which Jefferson posited as not only a right but a duty: "but when a long train of abuses and usurpations, begun at a distinguished period, and pursuing invariably the same object evinces a design to subject them to arbitrary power, it is their right, it is their duty, to throw off such government, and to provide new guards for their future security." https://www.let.rug.nl/usa/documents/1776-1785/jeffersons-draft-of-the-declaration-of-independence.php and that's what the Second Amendment if all about, it was designed as a bulwark against tyranny in government... Not having this right increases the "switching cost" to something out of range... And not only corps, but governments also are subject to enshittification...
Loved reading this. Not a programmer, but when I went (back) to grad school, a workstudy job was running the LAN for my dept, thru which I met a kindly gent running everything in the Dook dungeons. My job was boring and MG suggested I type "rn" at the *nix prompt. What a world! When I say the list of ahem newsgroups could be printed on about 10pp, you'll know it wasn't the beginning of the world but I could see it right behind & talk/listen to people who had a hand in building it.
Among others, I read alt.folklore.computers daily for years; frankly everyone I read conversed easily in these topics (so I understand more than I can do), and your piece took me right back to those days. Disclosure, big fan of esr, not least for his ability to explain things and willingness to change his mind, e.g., https://x.com/esrtweet/status/1910809356381413593 led to an epiphany of my own. In the thread you cite, perhaps you'd agree he implicitly acknowledges general correctness of what I take to be at least one main point of your comments (responding to https://x.com/CDoombeard/status/1982168304379728078). Anyway, thanks for the read.
I too think ESR is a great guy. He often has very incisive comments and indeed is willing to admit error, which is sadly rare these days
"in a large busy network of several dozen or more devices Ethernet became unusable"
I was at Xerox and then at 3Com. Both had very large numbers of peers on their Ethernet, were not "client-server" and "unusable" never happened. Please substantiate this. It's basically marketecture.
Novell COULD work on Ethernet but also on ARCNet, Corvus, or any other hardware people had. ARCNet was actually quite successful for a while.
Netware worked because it tended to break Ethernet into small segments that were routed. The same with TCP/IP once Cisco and Wellfleet released their first LAN routers. Small ethernet segments with almost all the traffic going to the router worked great. The catch was the router was more complex and expensive than a token ring bridge.
I don't know what the network at Xerox was like, but I do know that the Cambridge University implementation with lots of Xerox workstations turned out to be heavily server based as email and file storage was centralized. So yes the computers could talk peer to peer, but in practice 90% of the time they didn't. In fact once of the things Cambridge learned was that XServer sessions from one workstation to another broke things so much that they did ugly hacks like add a second NIC on a dedicated LAN for the workstations where people really needed to do that.
I expect 3com was the same. In fact IIRC 3com 3+ Open was a classic client-server system. Peer to peer sounds great but it turns out that except for chat apps most actual use cases are in fact client server.
The only thing I'd note about this was XServer (or X Windows) came much much later than the PARC Ethernet or 3+Open: late 80's or early 90's. In fact, calling a personal workstation a "server" seemed backwards to a lot of people, including me.
The times weren't that far apart. I know Cambridge had X in the late 1980s because among other things I had a project to make an Acorn Archimedes an XServer over a serial connection in 1988? (I think, maybe 89?) That was more or less the same time as I interned at Madge and reverse engineered a chunk of Novell's IPX protocol. That was the same time other Madge people were working with 3+Open and the Microsoft equivalent whose name escapes me.
XServer is definitely backwards though
"weren't that far apart"
far enough, though. By the late 80s most of the PARC heavyweights had departed for the three DEC labs in Palo Alto, and various other places.
Why do you think NetWare lost the great LAN wars?
I was doing contract work for Novell back in the day, writing up white papers and such for their product line. I went to Brainshare 2000, where there was a huge number of attendees. NetWare ruled the LANs.
Then NetWare 6 had its big release day on a particular Tuesday: 11 Sep 2001.
Bad omen, I guess. Then Microsoft came along and ate its lunch. Novell catered to the devs, but Microsoft wined and dined the suits.
It sounds like yet another example of the best tech not winning the day. Makes me sad.
Netware had issues from Netware 4 which took too long to get everything working properly with Win 95. That let Microsoft get the bugs out of NT Server and also let TCP/IP get a foot in the commercial door. Once MS put everything on IP, IPX was dead and that eventually killed the Novell model
First of all, they had Ray Noorda, who was old and had no obvious successor.
Then, they mostly ignored the Internet revolution.
Finally, Microsoft just outlasted them.
They might have had superiority in the 80s but they let it slip away.
I won't pretend to be a networking guru, but I've had to set up office networks in a few startups. One thing I learned early on was the technical differences between a "router", a "bridge", and a "hub".
Routers create a sort of network partition, and are smart enough to keep traffic inside their local network local, while forwarding traffic that needs to go out to the "outbound" network.
Hubs are just dumb repeaters. They often look like routers - and are a lot cheaper because they're dumber - but repeat traffic out to everything that's connected to them.
Bridges are basically bigger, more scalable, and more expensive routers that you need if you have multiple office networks that need to interact with each other.
Too many people would see a $15 hub and just fill their office network with them, and wonder why their office networking was so slow, especially when 10baseT was what most people used. A few times I pretended to be a networking guy just by figuring out where routers should be instead of hubs, buying a few $50 routers, deploying them, and suddenly the network is decently fast.
As for IBM worlds, the one thing that scaled better than IBM networks was IBM expense. A couple of startups I was at briefly chatted with IBM sales reps, got basically "the $2000 you are budgeting for your 50 device office network will be like $15K for an IBM network", and that was that.
Yes a few routers solved most Ethernet problems and IBM was never the choice for the budget conscious. That was why Token Ring failed IMO. These days the "you can buy better, but you can't pay more for it" flag has been handed off to Cisco
I have also seen it suggested that it was the high per-person license fees that prevented Smalltalk from flourishing.
Why pay $500/year when Perl and Python (and even certain versions of Common Lisp and Scheme) could be downloaded for free?
Well, I always like to see a non-expert guy solve a problem that Ph.D.'s and grad students struggle with and write whole textbooks on. Good for you.
IBM was used to charging ultra-premium prices, and counting on executives to say, "No one ever got fired for buying IBM." In the 80's that magic disappeared for good.
The Micro Channel (late 80's) was their attempt to kill the clones and regain control over computing. Amusingly enough, they paid 3Com (where I was) to develop an Ethernet adapter for it. So even then, they were realizing that the genie was out of the bottle.
Microchannel was yet another example of IBM trying to be sort of open standards but proprietary and screwing up as a result. MCA was objectively better than EISA but because they wouldn't license the bus to other PC makers it was destined to fail because IBM PS/2 servers were several times the price of almost as good Dells or Compaqs.
Needless to say a ton of the ideas in MCA were adopted by the subsequent PCI standard and IBM eventually ditched MCA for PCI.
I worked for DEC and we had Decent and email in 1985.....too bad Ken didn't believe in open systems
IBM was the same.
How would Token Ring work with a wireless PHY layer?
Like token bus. You just have a way to detect other members of the wireless LAN and hand the token from one to another.