3 ms·

Probably for the same reason we're still using that 3000-year-old state of the art known as "place-value": we haven't really found anything fundamentally better

Probably for the same reason we're still using that 3000-year-old state of the art known as "place-value": we haven't really found anything fundamentally better to replace it... yet.

asciilifeform 17y ago
> we haven't really found anything fundamentally better This is plainly false if you bother to lift a finger and read about the ideas of Genera, CoyotOS, ErOS, Smalltalk, etc. The UNIXes (and VMS, and its mutant Microsoft progeny) are pieces of junk which remain popular solely through network effects.
- bayareaguy 17y ago
  I have read about some of those. Unfortunately I must have missed or misunderstood something since I can't think of anything they offer that makes them worth adopting over Unix. It wouldn't be the first time something like this has happened. The idea of the World Wide Web wasn't all that impressive to me when I first saw it, partially because I'd already been using networks for almost a decade before it and was completely adapted to using arcane file transfer commands. It could be that there are some ideas in those systems that are just as revolutionary and I've missed them. Could you perhaps choose one or two of the ideas you think make those systems fundamentally better than Unix and elaborate on them?
  - asciilifeform 17y ago
    Orthogonal persistence, for one. All of the gymnastics we currently perform to make data outlive unplugging are a waste of precious lifetime: http://news.ycombinator.org/item?id=643794 http://news.ycombinator.org/item?id=643794
    - bayareaguy 17y ago
      Ok lets consider this. I'm willing to believe that we could eliminate some of these "gymnastics" by adopting new persistence models but not without paying a cost somewhere else. Perhaps that cost is only some additional education but I'm not convinced things are that simple. A decade or two ago many organizations tried proprietary object-oriented databases for various applications in part because they purported to simplify persistence issues. For some specific applications they did but more often than not the problems they caused their users were worse than the ones they solved and most ultimately decided to settle for app servers and ORM solutions. I've always thought a key barrier to the adoption of systems that support Orthogonal persistence is that it's harder to make applications designed around it perform as efficiently as ones designed to use lower level things like files or databases, partly because many techniques (e.g. efficient working set management, concurrency control, schema change, replication, data archival) are pretty well understood for applications that use the latter on systems like Unix. Maybe there are new opportunities for orthogonal persistence now that distributed cloud storage and computing is becoming popular. Such an OS specifically designed for application providers running on large clusters with S3-style storage could change things. However none of the systems you mentioned look like good candidates to me for such change.
      - asciilifeform 17y ago
        Forget databases. The machine gets an address space of 0..N bytes. They stay there, whether or not there is AC power. Internally this could consist of an array of mechanical or flash disks, cached by abundant cheap RAM. How difficult is that? No proprietary anything, no "object oriented" anything necessary. Nothing to think about. No changes in software at all, in fact, except for all of the various things you no longer have to do. You build a data structure in memory - doesn't matter what kind - and it stays there, until you erase it. That's what orthogonal persistence is. Forget the snake oils which try to coopt the phrase. One example of a current machine with actual orthogonal persistence is the Palm Pilot. > it's harder to make applications designed around it perform as efficiently Crank the R/W speed up to eleven and hope the issues go away. We're nearly there with striped solid-state disks. > OS specifically designed for application providers running on large clusters with S3-style storage Screw that, I just want a desktop which behaves sanely - from both the user and programmer's point of view. This means a single, nonvolatile address space.

Probably for the same reason we're still using that 3000-year-old state of the art known as "place-value": we haven't really found anything fundamentally better to replace it... yet.

asciilifeform 17y ago

> we haven't really found anything fundamentally better This is plainly false if you bother to lift a finger and read about the ideas of Genera, CoyotOS, ErOS, Smalltalk, etc. The UNIXes (and VMS, and its mutant Microsoft progeny) are pieces of junk which remain popular solely through network effects.

bayareaguy 17y ago
I have read about some of those. Unfortunately I must have missed or misunderstood something since I can't think of anything they offer that makes them worth adopting over Unix. It wouldn't be the first time something like this has happened. The idea of the World Wide Web wasn't all that impressive to me when I first saw it, partially because I'd already been using networks for almost a decade before it and was completely adapted to using arcane file transfer commands. It could be that there are some ideas in those systems that are just as revolutionary and I've missed them. Could you perhaps choose one or two of the ideas you think make those systems fundamentally better than Unix and elaborate on them?
- asciilifeform 17y ago
  Orthogonal persistence, for one. All of the gymnastics we currently perform to make data outlive unplugging are a waste of precious lifetime: http://news.ycombinator.org/item?id=643794 http://news.ycombinator.org/item?id=643794
  - bayareaguy 17y ago
    Ok lets consider this. I'm willing to believe that we could eliminate some of these "gymnastics" by adopting new persistence models but not without paying a cost somewhere else. Perhaps that cost is only some additional education but I'm not convinced things are that simple. A decade or two ago many organizations tried proprietary object-oriented databases for various applications in part because they purported to simplify persistence issues. For some specific applications they did but more often than not the problems they caused their users were worse than the ones they solved and most ultimately decided to settle for app servers and ORM solutions. I've always thought a key barrier to the adoption of systems that support Orthogonal persistence is that it's harder to make applications designed around it perform as efficiently as ones designed to use lower level things like files or databases, partly because many techniques (e.g. efficient working set management, concurrency control, schema change, replication, data archival) are pretty well understood for applications that use the latter on systems like Unix. Maybe there are new opportunities for orthogonal persistence now that distributed cloud storage and computing is becoming popular. Such an OS specifically designed for application providers running on large clusters with S3-style storage could change things. However none of the systems you mentioned look like good candidates to me for such change.
    - asciilifeform 17y ago
      Forget databases. The machine gets an address space of 0..N bytes. They stay there, whether or not there is AC power. Internally this could consist of an array of mechanical or flash disks, cached by abundant cheap RAM. How difficult is that? No proprietary anything, no "object oriented" anything necessary. Nothing to think about. No changes in software at all, in fact, except for all of the various things you no longer have to do. You build a data structure in memory - doesn't matter what kind - and it stays there, until you erase it. That's what orthogonal persistence is. Forget the snake oils which try to coopt the phrase. One example of a current machine with actual orthogonal persistence is the Palm Pilot. > it's harder to make applications designed around it perform as efficiently Crank the R/W speed up to eleven and hope the issues go away. We're nearly there with striped solid-state disks. > OS specifically designed for application providers running on large clusters with S3-style storage Screw that, I just want a desktop which behaves sanely - from both the user and programmer's point of view. This means a single, nonvolatile address space.