Skip to main content

View Diary: The MSFT-NSA conspiracy theory (40 comments)

Comment Preferences

  •  Here's a classic example (4+ / 0-)

    The man who wrote much of the Java run time was a professor before he went to work for Sun.  In his class on Software Defects, he would always start by asking someone to write a correct version of binary search.  He would always find bugs in the implementation; it turns out to be fairly subtle to get right.

    In fact, it was so subtle that the version he wrote for the Java RTL had a bug that persisted for five generations before it was reported to Sun:  http://en.wikipedia.org/....  Notice that this bug only manifested if one were performing a binary array search on an array of EXACTLY the maximum size on a 32 bit machine.  There was NO other way to evoke the bug.  You had to be running on a Pentium 5 with extended memory to see it, and you had to be running a huge Java program, and you had to hit exactly the right spot, and then, and only then, your program would fail incomprehensibly.

    Running for a long time doesn't prove the non-existence of defects.  All software has bugs.  Even the simplest software can have bugs.

    •  Proving a negative is always rather challenging (2+ / 0-)
      Recommended by:
      serendipityisabitch, kurt

      As is your assertion that ALL software has bugs.

      Not convinced.




      Somebody has to do something, and it's just incredibly pathetic that it has to be us.
      ~ Jerry Garcia

      by DeadHead on Sat Jun 15, 2013 at 09:06:05 PM PDT

      [ Parent ]

      •  True. I've been known to write a "Hello world" (1+ / 0-)
        Recommended by:
        DeadHead

        program that didn't have any. Less than 100 lines of code, probably okay. Between 100 and 1000, maybe. >100,000 lines and multiple programmers - um, er, well...

        At least half the future I've been expecting hasn't gotten here yet. Sigh.... (Yes, there's gender bias in my name; no, I wasn't thinking about it when I signed up. My apologies.)

        by serendipityisabitch on Sat Jun 15, 2013 at 09:21:48 PM PDT

        [ Parent ]

        •  I've dabbled in several languages on and off (1+ / 0-)
          Recommended by:
          Free Jazz at High Noon

          In a non-professional context since 1982.

          Apple BASIC, AutoLISP, C, VB, php, python, etc.

          I wrote programs that were bug free and performed exactly as expected. They were small scale, to be sure, but not every piece of software has bugs by default. And bug-free software is not unattainable. So this assertion that every program has bugs that have yet to be discovered is bullshit and impossible to verify.




          Somebody has to do something, and it's just incredibly pathetic that it has to be us.
          ~ Jerry Garcia

          by DeadHead on Sat Jun 15, 2013 at 09:32:40 PM PDT

          [ Parent ]

          •  Do you realize just how short binary search is? (4+ / 0-)

            The average routine is less than 10 lines long.  It's the kind of routine that one is expected to be able to write at the end of one's first programming language course.

            In the case of the languages you listed, you absolutely have never written a piece of code without a boatload of bugs.  The libraries themselves in all of those languages have known defects -- your program had bugs, you just never saw them.

            •  I see (1+ / 0-)
              Recommended by:
              Free Jazz at High Noon

              So my programs had bugs because the libraries used to compile them had bugs. Even though my programs did not produce erroneous results, it still had bugs, and, should a "perfect storm" of circumstances occur, my programs will then return erroneous results. Someday. Maybe. Not sure when, but someday.

              What do you recommend? Should we all just stop using Excel until we have certification from Microsoft that the compiler used to compile Excel was 100% defect free? Can we ever have confidence in our spreadsheet formulae returning accurate results?




              Somebody has to do something, and it's just incredibly pathetic that it has to be us.
              ~ Jerry Garcia

              by DeadHead on Sat Jun 15, 2013 at 11:01:59 PM PDT

              [ Parent ]

              •  What do I recommend? (0+ / 0-)

                To answer your last question first, it depends.  Technically, I trust Excel to about eight digits, but not much beyond that.  (At least as long as the code isn't running on a first-generation Pentium 4,)

                What I remember at all times is that the biggest vulnerability on my machine is the person providing input: me.    I never load any macros I haven't read and audited, and if they aren't obviously as-correct-as-possible to me, I use a different spreadsheet to export the content as csv and reimport it.

                To answer your second to last question: see above.  I use it, so I obviously consider the risks minimal.

Subscribe or Donate to support Daily Kos.

  • Recommended (122)
  • Community (57)
  • Memorial Day (31)
  • Culture (23)
  • Environment (20)
  • Rescued (20)
  • Law (20)
  • Civil Rights (20)
  • Science (18)
  • Marriage Equality (16)
  • Media (16)
  • Labor (16)
  • Elections (15)
  • Education (15)
  • Economy (14)
  • Republicans (14)
  • Trans-Pacific Partnership (14)
  • Ireland (14)
  • Josh Duggar (13)
  • Racism (12)
  • Click here for the mobile view of the site