3 ms·

I have no idea in truth but, if I had to guess, I'd say that this could be made transparent to existing applications using a decent abstraction layer in between

I have no idea in truth but, if I had to guess, I'd say that this could be made transparent to existing applications using a decent abstraction layer in between. You'd still pass something around that looks like a pointer, and you could use it the same way, but the OS would know it's not really. Of course, that's still potentially open to all kinds of abuse, not to mention a turtles-all-the-way-down counter-argument, so it really is just a guess.

Animats 9y ago
Apparently AmigaOS 4.1 (2008) doesn't have memory protection between processes, despite what the parent article says. No good way to make it backward compatible.[1] [1] https://arstechnica.com/information-technology/2008/09/amigaos41-ars/2/ https://arstechnica.com/information-technology/2008/09/amiga...
- bartread 9y ago
  Interesting, and also somewhat ironic that both articles are by the same guy. Either he forgot some of the limitations [1], or maybe there are additional improvements in the 4.1 release shipped with the X5000? [1] Granted, nearly 9 years have passed, so I don't think it's really fair to give him too hard a time for that.
- cat199 9y ago
  IIRC in the late pre-10 macos days, there was some kind of memory protection retrofitted to the platform, with some kind of permissions system allowing certain programs to break out of this when it really was required.. that could still presumably be done.. unless e.g. there are hardware restrictions (software expects processor to run in some 'mode', which it is not still in).. even then, that could probably be soft-faulted to an emulation shim in a modern full-featured CPU
  - Animats 9y ago
    The AmigaOS did, and does, interprocess communication by passing pointers across process boundaries. Access to the transmitted data is a simple memory access. The OS has no idea what data is supposed to be visible to the other process. The original MacOS didn't do that. Originally it only ran one program at a time, with some hacks to allow "desk accessories" to steal some cycles. The history of computing might have been quite different if the Motorola 68000 had correct instruction continuation after a memory access fault. (This was fixed in the 68010, 3 years later.) Then an MMU would have been possible sooner. Motorola was very late in producing a good MMU for the 68K line, and it wasn't until 1990 that they had an on-chip MMU.
    - eschaton 9y ago
      The 68030 had a great on-chip MMU in 1987, three years after the 68020 and its off-chip 68851 (which added a wait state). You're thinking of the 68040, which brought the FPU on-chip as well.
    - wolfgke 9y ago
      > The history of computing might have been quite different if the Motorola 68000 had correct instruction continuation after a memory access fault. (This was fixed in the 68010, 3 years later.) Can you give more details, in particular on details of the problem and how it was solved?

Animats 9y ago

Apparently AmigaOS 4.1 (2008) doesn't have memory protection between processes, despite what the parent article says. No good way to make it backward compatible.[1] [1] https://arstechnica.com/information-technology/2008/09/amigaos41-ars/2/ https://arstechnica.com/information-technology/2008/09/amiga...

bartread 9y ago
Interesting, and also somewhat ironic that both articles are by the same guy. Either he forgot some of the limitations [1], or maybe there are additional improvements in the 4.1 release shipped with the X5000? [1] Granted, nearly 9 years have passed, so I don't think it's really fair to give him too hard a time for that.
cat199 9y ago
IIRC in the late pre-10 macos days, there was some kind of memory protection retrofitted to the platform, with some kind of permissions system allowing certain programs to break out of this when it really was required.. that could still presumably be done.. unless e.g. there are hardware restrictions (software expects processor to run in some 'mode', which it is not still in).. even then, that could probably be soft-faulted to an emulation shim in a modern full-featured CPU
- Animats 9y ago
  The AmigaOS did, and does, interprocess communication by passing pointers across process boundaries. Access to the transmitted data is a simple memory access. The OS has no idea what data is supposed to be visible to the other process. The original MacOS didn't do that. Originally it only ran one program at a time, with some hacks to allow "desk accessories" to steal some cycles. The history of computing might have been quite different if the Motorola 68000 had correct instruction continuation after a memory access fault. (This was fixed in the 68010, 3 years later.) Then an MMU would have been possible sooner. Motorola was very late in producing a good MMU for the 68K line, and it wasn't until 1990 that they had an on-chip MMU.
  - eschaton 9y ago
    The 68030 had a great on-chip MMU in 1987, three years after the 68020 and its off-chip 68851 (which added a wait state). You're thinking of the 68040, which brought the FPU on-chip as well.
  - wolfgke 9y ago
    > The history of computing might have been quite different if the Motorola 68000 had correct instruction continuation after a memory access fault. (This was fixed in the 68010, 3 years later.) Can you give more details, in particular on details of the problem and how it was solved?