That's fascinating that less were implementation errors than design errors, though "the distribution of errors among families of algorithms is not uniform" accounts for that I guess. Though the paper is from 1983...
I scanned through the paper and found no explicit mention of the typical coding error caused by using M := (L+H)/2 instead of M := L + ((H - L) / 2) (though the paper interchanges both). So I suspect a re-analysis would find more coding errors than design errors in languages without arbitrary-sized integer auto-conversions. My reasoning for that conclusion is based on: http://googleresearch.blogspot.com/2006/06/extra-extra-read-...
Of course, given the stories of all the CS-degree wielding applicants who flat-out can't do FizzBuzz, and the amusing incorrect designs/implementations of commenters who scoff at the notion of not being able to do FizzBuzz and try to prove they can but fail, I was already inclined to believe there's a lot of incompetence to go around. It's not even necessarily a problem with the CS programs, but there is the danger of learning to recite an algorithm's steps word for word without knowing what it means or what it's useful for and fail with a live use case.
I scanned through the paper and found no explicit mention of the typical coding error caused by using M := (L+H)/2 instead of M := L + ((H - L) / 2) (though the paper interchanges both). So I suspect a re-analysis would find more coding errors than design errors in languages without arbitrary-sized integer auto-conversions. My reasoning for that conclusion is based on: http://googleresearch.blogspot.com/2006/06/extra-extra-read-...
Of course, given the stories of all the CS-degree wielding applicants who flat-out can't do FizzBuzz, and the amusing incorrect designs/implementations of commenters who scoff at the notion of not being able to do FizzBuzz and try to prove they can but fail, I was already inclined to believe there's a lot of incompetence to go around. It's not even necessarily a problem with the CS programs, but there is the danger of learning to recite an algorithm's steps word for word without knowing what it means or what it's useful for and fail with a live use case.