3 ms·

> Dijkstra will go down in history as one of the great algorithm designers, but this will always be a black mark on his legacy, possibly overshading his contrib

> Dijkstra will go down in history as one of the great algorithm designers, but this will always be a black mark on his legacy, possibly overshading his contributions. There's a lesson here about not pontificating outside of your field I am begging of you, please do a little bit of investigation into the historical context in which Dijkstra wrote that memo.

phendrenad2 3y ago
Your comment was likely inspired by other replies to my comment, but please, don't assume someone who has an opinion other than yours simply hasn't done enough research to agree with you.
- bigbillheck 3y ago
  In one 1971 dataset fully 13% of all executable statements were "GOTO": https://www.cs.tufts.edu/~nr/cs257/archive/don-knuth/empirical-fortran.pdf https://www.cs.tufts.edu/~nr/cs257/archive/don-knuth/empiric...
  - phendrenad2 3y ago
    In 1971 people really cared about binary density.
    - bigbillheck 3y ago
      Reading that paper it's hard to believe that the typical programmer cared all that much beyond getting something to work. Whatever you mean by "binary density", the subjects were programming for an IBM Model 67 which was a pretty hefty system for the time. The programs weren't all that big by modern standards either, with "number of executable statements" in the hundreds; when you have megabytes of memory it doesn't seem like spending a few extra bytes on instructions would be that big a deal. By and large the ancients couldn't have been all that concerned about optimizations. Take a look at their "Example 6" in which the authors were able to significantly improve an inner loop thru such advanced techniques as "use 0.0 instead of 0 if you want a REAL" and "don't make extra assignments".
      - phendrenad2 3y ago
        Do you have some evidence to back any of this up? You're claiming that software developers who had access to an IBM Model 67 didn't actually make use of its full storage space. In that case, why not buy a smaller machine? It sounds completely fantastical/fabricated to me.
        bigbillheck 3y ago
        > Do you have some evidence to back any of this up? The paper that I linked? > In that case, why not buy a smaller machine? Because they needed to support an entire department or university worth of users, many of them at the same time. > It sounds completely fantastical/fabricated to me. Have you ever done serious work on a time-shared system? The kind where you type `w` and see a bunch of people who aren't you? Nowadays everybody gets their own workstation and can use as much or as little of it as they see fit, but on a shared system trying to use the whole thing is a great way to get people mad at you, because you're taking away resources they need for their own work. Not that you'd necessarily be allowed to use all that much in the first place, that's why things like `ulimit` exist.
        phendrenad2 3y ago
        Sounds like you're actually agreeing with me. If there are multiple users, then that just puts more pressure on users to write compact code.

phendrenad2 3y ago

Your comment was likely inspired by other replies to my comment, but please, don't assume someone who has an opinion other than yours simply hasn't done enough research to agree with you.

bigbillheck 3y ago
In one 1971 dataset fully 13% of all executable statements were "GOTO": https://www.cs.tufts.edu/~nr/cs257/archive/don-knuth/empirical-fortran.pdf https://www.cs.tufts.edu/~nr/cs257/archive/don-knuth/empiric...
- phendrenad2 3y ago
  In 1971 people really cared about binary density.
  - bigbillheck 3y ago
    Reading that paper it's hard to believe that the typical programmer cared all that much beyond getting something to work. Whatever you mean by "binary density", the subjects were programming for an IBM Model 67 which was a pretty hefty system for the time. The programs weren't all that big by modern standards either, with "number of executable statements" in the hundreds; when you have megabytes of memory it doesn't seem like spending a few extra bytes on instructions would be that big a deal. By and large the ancients couldn't have been all that concerned about optimizations. Take a look at their "Example 6" in which the authors were able to significantly improve an inner loop thru such advanced techniques as "use 0.0 instead of 0 if you want a REAL" and "don't make extra assignments".
    - phendrenad2 3y ago
      Do you have some evidence to back any of this up? You're claiming that software developers who had access to an IBM Model 67 didn't actually make use of its full storage space. In that case, why not buy a smaller machine? It sounds completely fantastical/fabricated to me.
      - bigbillheck 3y ago
        > Do you have some evidence to back any of this up? The paper that I linked? > In that case, why not buy a smaller machine? Because they needed to support an entire department or university worth of users, many of them at the same time. > It sounds completely fantastical/fabricated to me. Have you ever done serious work on a time-shared system? The kind where you type `w` and see a bunch of people who aren't you? Nowadays everybody gets their own workstation and can use as much or as little of it as they see fit, but on a shared system trying to use the whole thing is a great way to get people mad at you, because you're taking away resources they need for their own work. Not that you'd necessarily be allowed to use all that much in the first place, that's why things like `ulimit` exist.
        phendrenad2 3y ago
        Sounds like you're actually agreeing with me. If there are multiple users, then that just puts more pressure on users to write compact code.