4 ms·

> mystique would be lost People would also discover that the architecture is not simple, but grown over a decade while there was rapid development of the basic

by lispm 3y ago

> mystique would be lost People would also discover that the architecture is not simple, but grown over a decade while there was rapid development of the basics people take for granted: standards, languages, networking, operating systems, ... In the mid 80s TCP/IP was an expensive add on for Genera. Later it was made a part of the OS. But development stopped before it could add things like IPv6, Unicode, basic security, HTTPS, and other things. One could add it, but then one had to learn ZetaLisp (70s/80s), Flavors (early 80s), Symbolics Common Lisp (80s/90s), an object-oriented networking stack architecture from the 80s, ... You would have a second life in a technology stack somewhere between the past and the future, in a parallel world.

cmrdporcupine 3y ago
It makes more sense to me for some group to attempt to replicate the IDE experience in a modern Lisp (or similar) flavour. Still I also think these kinds of "live editing" environments don't translate well into some modern "best practices" around version control, deployment, versioning, etc. Remembering my experience with LambdaMOO & similar and then later Smalltalk/Squeak and recently with doing stuff with Julia's REPL, etc. They're fascinating and fairly productive, but I am not convinced about this technique's ability to scale.
- lispm 3y ago
  > It makes more sense to me for some group to attempt to replicate the IDE experience in a modern Lisp (or similar) flavour. It may not have the fancy UI, may not have a fancy user experience, may not be an operating system, may have a more primitive Lisp dialect, may not have a good multi-threading story, ... but it might exist and people put a lot of work into it: GNU Emacs and its extensions. That's how it is...
  - cmrdporcupine 3y ago
    I think half the magic I associate with the Symbolics products is that they also made their own hardware, custom built for Lisp. Which is kinda neat and magical, if entirely impractical. That plus the lovely keyboards, nice case design, etc. etc.
    - lispm 3y ago
      In the early days they had to - all the hardware was non-standard. Where would one get 36 bit memory cards? Where would one get CPUs which had a Lisp-specific instruction set? In those early days the computer (refrigerator size) would sit in a machine room (with enough power) and the user would be in his/her room with only a console (plus maybe a second monitor), keyboard and mouse - connected via a long console cable to the machine in the machine room. That was the experience for the programmer/user. The machine itself could have a lot of peripherals: network, tape drive, memory boards, color boards, frame grabber, disk drives, cpu accelerator, ... the programmer would see the driver code. At some point in time they produced cards for the SUN VMEBus and for the Mac Nubus. Then the only thing left was the keyboard. So, I agree, the Lisp Machine concept was a combination of Hard- and Software. The emulators of today only provide some of the software parts...
      - cmrdporcupine 3y ago
        Yeah I have here the book published by the folks who do did the Linn Rekursiv computer, which is sorta like the same philosophy as the Lisp Machine but for a highly object oriented system. Tagged memory, hardware supported garbage collection, object-structured memory, etc. I believe it was also positioned as a VMEbus card, etc. My understanding is Moore's law just ended up making these attempts uneconomical. By the time HW eng and manufacturing is done on the custom components, "orthodox" general MPU capabilities end up leapfrogging it and a well engineered software VM can outperform. Maybe we could start to see this change again, who knows. Maybe unikernels running highly tuned VMs on the hypervisor could be this generation's equivalent. Full control over the MMU, tagged pointers all the way down... hmmmmm
        lispm 3y ago
        Symbolics could have built computers with custom hardware. At some point in time the 64bit RISC CPUs (MIPS, Alpha, SPARC, POWER, x86-64, ...) were all powerful enough to run Lisp. Genera got ported to the DEC Alpha). They would have been also powerful enough to run Lisp as an OS. But that would also mean writing device drivers, interfaces to hardware and a lot of other stuff. It would have cost substantial amounts of money to port Genera to the metal. Who would have bought that? The idea of a high-level language running as an OS on the metal did not have any future. I don't think there are good examples where this was successful on the market.

by lispm 3y ago

cmrdporcupine 3y ago

It makes more sense to me for some group to attempt to replicate the IDE experience in a modern Lisp (or similar) flavour. Still I also think these kinds of "live editing" environments don't translate well into some modern "best practices" around version control, deployment, versioning, etc. Remembering my experience with LambdaMOO & similar and then later Smalltalk/Squeak and recently with doing stuff with Julia's REPL, etc. They're fascinating and fairly productive, but I am not convinced about this technique's ability to scale.

lispm 3y ago
> It makes more sense to me for some group to attempt to replicate the IDE experience in a modern Lisp (or similar) flavour. It may not have the fancy UI, may not have a fancy user experience, may not be an operating system, may have a more primitive Lisp dialect, may not have a good multi-threading story, ... but it might exist and people put a lot of work into it: GNU Emacs and its extensions. That's how it is...
- cmrdporcupine 3y ago
  I think half the magic I associate with the Symbolics products is that they also made their own hardware, custom built for Lisp. Which is kinda neat and magical, if entirely impractical. That plus the lovely keyboards, nice case design, etc. etc.
  - lispm 3y ago
    In the early days they had to - all the hardware was non-standard. Where would one get 36 bit memory cards? Where would one get CPUs which had a Lisp-specific instruction set? In those early days the computer (refrigerator size) would sit in a machine room (with enough power) and the user would be in his/her room with only a console (plus maybe a second monitor), keyboard and mouse - connected via a long console cable to the machine in the machine room. That was the experience for the programmer/user. The machine itself could have a lot of peripherals: network, tape drive, memory boards, color boards, frame grabber, disk drives, cpu accelerator, ... the programmer would see the driver code. At some point in time they produced cards for the SUN VMEBus and for the Mac Nubus. Then the only thing left was the keyboard. So, I agree, the Lisp Machine concept was a combination of Hard- and Software. The emulators of today only provide some of the software parts...
    - cmrdporcupine 3y ago
      Yeah I have here the book published by the folks who do did the Linn Rekursiv computer, which is sorta like the same philosophy as the Lisp Machine but for a highly object oriented system. Tagged memory, hardware supported garbage collection, object-structured memory, etc. I believe it was also positioned as a VMEbus card, etc. My understanding is Moore's law just ended up making these attempts uneconomical. By the time HW eng and manufacturing is done on the custom components, "orthodox" general MPU capabilities end up leapfrogging it and a well engineered software VM can outperform. Maybe we could start to see this change again, who knows. Maybe unikernels running highly tuned VMs on the hypervisor could be this generation's equivalent. Full control over the MMU, tagged pointers all the way down... hmmmmm
      - lispm 3y ago
        Symbolics could have built computers with custom hardware. At some point in time the 64bit RISC CPUs (MIPS, Alpha, SPARC, POWER, x86-64, ...) were all powerful enough to run Lisp. Genera got ported to the DEC Alpha). They would have been also powerful enough to run Lisp as an OS. But that would also mean writing device drivers, interfaces to hardware and a lot of other stuff. It would have cost substantial amounts of money to port Genera to the metal. Who would have bought that? The idea of a high-level language running as an OS on the metal did not have any future. I don't think there are good examples where this was successful on the market.