![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
My contention is that a large part of the reason that we have the crappy computers that we do today -- lowest-common-denominator boxes, mostly powered by one of the kludgiest and most inelegant CPU architectures of the last 40 years -- is not technical, nor even primarily commercial or due to business pressures, but rather, it's cultural.
When I was playing with home micros (mainly Sinclair and Amstrad; the American stuff was just too expensive for Brits in the early-to-mid 1980s), the culture was that Real Men programmed in assembler and the main battle was Z80 versus 6502, with a few weirdos saying that 6809 was better than either. BASIC was the language for beginners, and a few weirdos maintained that Forth was better.
At university, I used a VAXcluster and learned to program in Fortran-77. The labs had Acorn BBC Micros in -- solid machines, the best 8-bit BASIC ever, and they could interface both with lab equipment over IEEE-488 and with generic printers and so on over Centronics parallel and its RS-423 interface [EDIT: fixed!], which could talk to RS-232 kit.
As I discovered when I moved into the professional field a few years later (1988), this wasn't that different from the pro stuff. A lot of apps were written in various BASICs, and in the old era of proprietary OSes on proprietary kit, for performance, you used assembler.
But a new wave was coming. MS-DOS was already huge and the Mac was growing strongly. Windows was on v2 and was a toy, but Unix was coming to mainstream kit, or at least affordable kit. You could run Unix on PCs (e.g. SCO Xenix), on Macs (A/UX), and my employers had a demo IBM RT-6150 running AIX 1.
Unix wasn't only the domain (pun intentional) of expensive kit priced in the tens of thousands.
A new belief started to spread: that if you used C, you could get near-assembler performance without the pain, and the code could be ported between machines. DOS and Mac apps started to be written (or rewritten) in C, and some were even ported to Xenix. In my world, nobody used stuff like A/UX or AIX, and Xenix was specialised. I was aware of Coherent as the only "affordable" Unix, but I never saw a copy or saw it running.
So this second culture of C code running on non-Unix OSes appeared. Then the OSes started to scramble to catch up with Unix -- first OS/2, then Windows 3, then the for a decade parallel universe of Windows NT, until XP became established and Win9x finally died. Meanwhile, Apple and IBM flailed around, until IBM surrendered, Apple merged with NeXT and switched to NeXTstep.
Now, Windows is evolving to be more and more Unix-like, with GUI-less versions, clean(ish) separation between GUI and console apps, a new rich programmable shell, and so on.
While the Mac is now a Unix box, albeit a weird one.
Commercial Unix continues to wither away. OpenVMS might make a modest comeback. IBM mainframes seem to be thriving; every other kind of big iron is now emulated on x86 kit, as far as I can tell. IBM has successfully killed off several efforts to do this for z Series.
So now, it's Unix except for the single remaining mainstream proprietary system: Windows. Unix today means Linux, while the weirdoes use FreeBSD. Everything else seems to be more or less a rounding error.
C always was like carrying water in a sieve, so now, we have multiple C derivatives, trying to patch the holes. C++ has grown up but it's like Ada now: so huge that nobody understands it all, but actually, a fairly usable tool.
There's the kinda-sorta FOSS "safe C++ in a VM", Java. The proprietary kinda-sorta "safe C++ in a VM", C#. There's the not-remotely-safe kinda-sorta C in a web browser, Javascript.
And dozens of others, of course.
Even the safer ones run on a basis of C -- so the lovely cuddly friendly Python, that everyone loves, has weird C printing semantics to mess up the heads of beginners.
Perl has abandoned its base, planned to move onto a VM, then the VM went wrong, and now has a new VM and to general amazement and lack of interest, Perl 6 is finally here.
All the others are still implemented in C, mostly on a Unix base, like Ruby, or on a JVM base, like Clojure and Scala.
So they still have C like holes and there are frequent patches and updates to try to make them able to retain some water for a short time, while the "cyber criminals" make hundreds of millions.
Anything else is "uncommercial" or "not viable for real world use".
Borland totally dropped the ball and lost a nice little earner in Delphi, but it continues as Free Pascal and so on.
Apple goes its own way, but has forgotten the truly innovative projects it had pre-NeXT, such as Dylan.
There were real projects that were actually used for real work, like Oberon the OS, written in Oberon the language. Real pioneering work in UIs, such as Jef Raskin's machines, the original Mac and Canon Cat -- forgotten. People rhapsodise over the Amiga and forget that the planned OS, CAOS, to be as radical as the hardware, never made it out of the lab. Same, on a smaller scale, with the Acorn Archimedes.
Despite that, of course, Lisp never went away. People still use it, but they keep their heads down and get on with it.
Much the same applies to Smalltalk. Still there, still in use, still making real money and doing real work, but forgotten all the same.
The Lisp Machines and Smalltalk boxes lost the workstation war. Unix won, and as history is written by the victors, now the alternatives are forgotten or dismissed as weird kooky toys of no serious merit.
The senior Apple people didn't understand the essence of what they saw at PARC: they only saw the chrome. They copied the chrome, not the essence, and now all that any of us have is the chrome. We have GUIs, but on top of the nasty kludgy hacks of C and the like. A late-'60s skunkware project now runs the world, and the real serious research efforts to make something better, both before and after, are forgotten historical footnotes.
Modern computers are a vast disappointment to me. We have no thinking machines. The Fifth Generation, Lisp, all that -- gone.
What did we get instead?
Like dinosaurs, the expensive high-end machines of the '70s and '80s didn't evolve into their successors. They were just replaced. First little cheapo 8-bits, not real or serious at all, although they were cheap and people did serious stuff with them because it's all they could afford. The early 8-bits ran semi-serious OSes such as CP/M, but when their descendants sold a thousand times more, those descendants weren't running descendants of that OS -- no, it and its creator died.
CP/M evolved into a multiuser multitasking 386 OS that could run multiple MS-DOS apps on terminals, but it died.
No, then the cheapo 8-bits thrived in the form of an 8/16-bit hybrid, the 8086 and 8088, and a cheapo knock-off of CP/M.
This got a redesign into something grown-up: OS/2.
Predictably, that died.
So the hacked-together GUI for DOS got re-invigorated with an injection of OS/2 code, as Windows 3. That took over the world.
The rivals - the Amiga, ST, etc? 680x0 chips, lots of flat memory, whizzy graphics and sound? All dead.
Then Windows got re-invented with some OS/2 3 ideas and code, and some from VMS, and we got Windows NT.
But the marketing men got to it and ruined its security and elegance, to produce the lipstick-and-high-heels Windows XP. That version, insecure and flakey with its terrible bodged-in browser, that, of course, was the one that sold.
Linux got nowhere until it copied the XP model. The days of small programs, everything's a text file, etc. -- all forgotten. Nope, lumbering GUI apps, CORBA and RPC and other weird plumbing, huge complex systems, but it looks and works kinda like Windows and a Mac now so it looks like them and people use it.
Android looks kinda like iOS and people use it in their billions. Newton? Forgotten. No, people have Unix in their pocket, only it's a bloated successor of Unix.
The efforts to fix and improve Unix -- Plan 9, Inferno -- forgotten. A proprietary microkernel Unix-like OS for phones -- Blackberry 10, based on QNX -- not Androidy enough, and bombed.
We have less and less choice, made from worse parts on worse foundations -- but it's colourful and shiny and the world loves it.
That makes me despair.
We have poor-quality tools, built on poorly-designed OSes, running on poorly-designed chips. Occasionally, fragments of older better ways, such as functional-programming tools, or Lisp-based development environments, are layered on top of them, but while they're useful in their way, they can't fix the real problems underneath.
Occasionally someone comes along and points this out and shows a better way -- such as Curtis Yarvin's Urbit. Lisp Machines re-imagined for the 21st century, based on top of modern machines. But nobody gets it, and its programmer has some unpleasant and unpalatable ideas, so it's doomed.
And the kids who grew up after C won the battle deride the former glories, the near-forgotten brilliance that we have lost.
And it almost makes me want to cry sometimes.
We should have brilliant machines now, not merely Steve Jobs' "bicycles for the mind", but Gossamer Albatross-style hang-gliders for the mind.
But we don't. We have glorified 8-bits. They multitask semi-reliably, they can handle sound and video and 3D and look pretty. On them, layered over all the rubbish and clutter and bodges and hacks, inspired kids are slowly brute-forcing machines that understand speech, which can see and walk and drive.
But it could have been so much better.
Charles Babbage didn't finish the Difference Engine. It would have paid for him to build his Analytical Engine, and that would have given the Victorian British Empire the steam-driven computer, which would have transformed history.
But he got distracted and didn't deliver.
We started to build what a few old-timers remember as brilliant machines, machines that helped their users to think and to code, with brilliant -- if flawed -- software written in the most sophisticated computer languages yet devised, by the popular acclaim of the people who really know this stuff: Lisp and Smalltalk.
But we didn't pursue them. We replaced them with something cheaper -- with Unix machines, an OS only a nerd could love. And then we replaced the Unix machines with something cheaper still -- the IBM PC, a machine so poor that the £125 ZX Spectrum had better graphics and sound.
And now, we all use descendants of that. Generally acknowledged as one of the poorest, most-compromised machines, based on descendants of one of the poorest, most-compromised CPUs.
Yes, over the 40 years since then, most of rough edges have been polished out. The machines are now small, fast, power-frugal with tons of memory and storage, with great graphics and sound. But it's taken decades to get here.
And the OSes have developed. Now they're feature-rich, fairly friendly, really very robust considering the stone-age stuff they're built from.
But if we hadn't spent 3 or 4 decades making a pig's ear into silk purse -- if we'd started with a silk purse instead -- where might we have got to by now?
When I was playing with home micros (mainly Sinclair and Amstrad; the American stuff was just too expensive for Brits in the early-to-mid 1980s), the culture was that Real Men programmed in assembler and the main battle was Z80 versus 6502, with a few weirdos saying that 6809 was better than either. BASIC was the language for beginners, and a few weirdos maintained that Forth was better.
At university, I used a VAXcluster and learned to program in Fortran-77. The labs had Acorn BBC Micros in -- solid machines, the best 8-bit BASIC ever, and they could interface both with lab equipment over IEEE-488 and with generic printers and so on over Centronics parallel and its RS-423 interface [EDIT: fixed!], which could talk to RS-232 kit.
As I discovered when I moved into the professional field a few years later (1988), this wasn't that different from the pro stuff. A lot of apps were written in various BASICs, and in the old era of proprietary OSes on proprietary kit, for performance, you used assembler.
But a new wave was coming. MS-DOS was already huge and the Mac was growing strongly. Windows was on v2 and was a toy, but Unix was coming to mainstream kit, or at least affordable kit. You could run Unix on PCs (e.g. SCO Xenix), on Macs (A/UX), and my employers had a demo IBM RT-6150 running AIX 1.
Unix wasn't only the domain (pun intentional) of expensive kit priced in the tens of thousands.
A new belief started to spread: that if you used C, you could get near-assembler performance without the pain, and the code could be ported between machines. DOS and Mac apps started to be written (or rewritten) in C, and some were even ported to Xenix. In my world, nobody used stuff like A/UX or AIX, and Xenix was specialised. I was aware of Coherent as the only "affordable" Unix, but I never saw a copy or saw it running.
So this second culture of C code running on non-Unix OSes appeared. Then the OSes started to scramble to catch up with Unix -- first OS/2, then Windows 3, then the for a decade parallel universe of Windows NT, until XP became established and Win9x finally died. Meanwhile, Apple and IBM flailed around, until IBM surrendered, Apple merged with NeXT and switched to NeXTstep.
Now, Windows is evolving to be more and more Unix-like, with GUI-less versions, clean(ish) separation between GUI and console apps, a new rich programmable shell, and so on.
While the Mac is now a Unix box, albeit a weird one.
Commercial Unix continues to wither away. OpenVMS might make a modest comeback. IBM mainframes seem to be thriving; every other kind of big iron is now emulated on x86 kit, as far as I can tell. IBM has successfully killed off several efforts to do this for z Series.
So now, it's Unix except for the single remaining mainstream proprietary system: Windows. Unix today means Linux, while the weirdoes use FreeBSD. Everything else seems to be more or less a rounding error.
C always was like carrying water in a sieve, so now, we have multiple C derivatives, trying to patch the holes. C++ has grown up but it's like Ada now: so huge that nobody understands it all, but actually, a fairly usable tool.
There's the kinda-sorta FOSS "safe C++ in a VM", Java. The proprietary kinda-sorta "safe C++ in a VM", C#. There's the not-remotely-safe kinda-sorta C in a web browser, Javascript.
And dozens of others, of course.
Even the safer ones run on a basis of C -- so the lovely cuddly friendly Python, that everyone loves, has weird C printing semantics to mess up the heads of beginners.
Perl has abandoned its base, planned to move onto a VM, then the VM went wrong, and now has a new VM and to general amazement and lack of interest, Perl 6 is finally here.
All the others are still implemented in C, mostly on a Unix base, like Ruby, or on a JVM base, like Clojure and Scala.
So they still have C like holes and there are frequent patches and updates to try to make them able to retain some water for a short time, while the "cyber criminals" make hundreds of millions.
Anything else is "uncommercial" or "not viable for real world use".
Borland totally dropped the ball and lost a nice little earner in Delphi, but it continues as Free Pascal and so on.
Apple goes its own way, but has forgotten the truly innovative projects it had pre-NeXT, such as Dylan.
There were real projects that were actually used for real work, like Oberon the OS, written in Oberon the language. Real pioneering work in UIs, such as Jef Raskin's machines, the original Mac and Canon Cat -- forgotten. People rhapsodise over the Amiga and forget that the planned OS, CAOS, to be as radical as the hardware, never made it out of the lab. Same, on a smaller scale, with the Acorn Archimedes.
Despite that, of course, Lisp never went away. People still use it, but they keep their heads down and get on with it.
Much the same applies to Smalltalk. Still there, still in use, still making real money and doing real work, but forgotten all the same.
The Lisp Machines and Smalltalk boxes lost the workstation war. Unix won, and as history is written by the victors, now the alternatives are forgotten or dismissed as weird kooky toys of no serious merit.
The senior Apple people didn't understand the essence of what they saw at PARC: they only saw the chrome. They copied the chrome, not the essence, and now all that any of us have is the chrome. We have GUIs, but on top of the nasty kludgy hacks of C and the like. A late-'60s skunkware project now runs the world, and the real serious research efforts to make something better, both before and after, are forgotten historical footnotes.
Modern computers are a vast disappointment to me. We have no thinking machines. The Fifth Generation, Lisp, all that -- gone.
What did we get instead?
Like dinosaurs, the expensive high-end machines of the '70s and '80s didn't evolve into their successors. They were just replaced. First little cheapo 8-bits, not real or serious at all, although they were cheap and people did serious stuff with them because it's all they could afford. The early 8-bits ran semi-serious OSes such as CP/M, but when their descendants sold a thousand times more, those descendants weren't running descendants of that OS -- no, it and its creator died.
CP/M evolved into a multiuser multitasking 386 OS that could run multiple MS-DOS apps on terminals, but it died.
No, then the cheapo 8-bits thrived in the form of an 8/16-bit hybrid, the 8086 and 8088, and a cheapo knock-off of CP/M.
This got a redesign into something grown-up: OS/2.
Predictably, that died.
So the hacked-together GUI for DOS got re-invigorated with an injection of OS/2 code, as Windows 3. That took over the world.
The rivals - the Amiga, ST, etc? 680x0 chips, lots of flat memory, whizzy graphics and sound? All dead.
Then Windows got re-invented with some OS/2 3 ideas and code, and some from VMS, and we got Windows NT.
But the marketing men got to it and ruined its security and elegance, to produce the lipstick-and-high-heels Windows XP. That version, insecure and flakey with its terrible bodged-in browser, that, of course, was the one that sold.
Linux got nowhere until it copied the XP model. The days of small programs, everything's a text file, etc. -- all forgotten. Nope, lumbering GUI apps, CORBA and RPC and other weird plumbing, huge complex systems, but it looks and works kinda like Windows and a Mac now so it looks like them and people use it.
Android looks kinda like iOS and people use it in their billions. Newton? Forgotten. No, people have Unix in their pocket, only it's a bloated successor of Unix.
The efforts to fix and improve Unix -- Plan 9, Inferno -- forgotten. A proprietary microkernel Unix-like OS for phones -- Blackberry 10, based on QNX -- not Androidy enough, and bombed.
We have less and less choice, made from worse parts on worse foundations -- but it's colourful and shiny and the world loves it.
That makes me despair.
We have poor-quality tools, built on poorly-designed OSes, running on poorly-designed chips. Occasionally, fragments of older better ways, such as functional-programming tools, or Lisp-based development environments, are layered on top of them, but while they're useful in their way, they can't fix the real problems underneath.
Occasionally someone comes along and points this out and shows a better way -- such as Curtis Yarvin's Urbit. Lisp Machines re-imagined for the 21st century, based on top of modern machines. But nobody gets it, and its programmer has some unpleasant and unpalatable ideas, so it's doomed.
And the kids who grew up after C won the battle deride the former glories, the near-forgotten brilliance that we have lost.
And it almost makes me want to cry sometimes.
We should have brilliant machines now, not merely Steve Jobs' "bicycles for the mind", but Gossamer Albatross-style hang-gliders for the mind.
But we don't. We have glorified 8-bits. They multitask semi-reliably, they can handle sound and video and 3D and look pretty. On them, layered over all the rubbish and clutter and bodges and hacks, inspired kids are slowly brute-forcing machines that understand speech, which can see and walk and drive.
But it could have been so much better.
Charles Babbage didn't finish the Difference Engine. It would have paid for him to build his Analytical Engine, and that would have given the Victorian British Empire the steam-driven computer, which would have transformed history.
But he got distracted and didn't deliver.
We started to build what a few old-timers remember as brilliant machines, machines that helped their users to think and to code, with brilliant -- if flawed -- software written in the most sophisticated computer languages yet devised, by the popular acclaim of the people who really know this stuff: Lisp and Smalltalk.
But we didn't pursue them. We replaced them with something cheaper -- with Unix machines, an OS only a nerd could love. And then we replaced the Unix machines with something cheaper still -- the IBM PC, a machine so poor that the £125 ZX Spectrum had better graphics and sound.
And now, we all use descendants of that. Generally acknowledged as one of the poorest, most-compromised machines, based on descendants of one of the poorest, most-compromised CPUs.
Yes, over the 40 years since then, most of rough edges have been polished out. The machines are now small, fast, power-frugal with tons of memory and storage, with great graphics and sound. But it's taken decades to get here.
And the OSes have developed. Now they're feature-rich, fairly friendly, really very robust considering the stone-age stuff they're built from.
But if we hadn't spent 3 or 4 decades making a pig's ear into silk purse -- if we'd started with a silk purse instead -- where might we have got to by now?
