I think the name should be changed, since it's not dependent on experimentation, and largely has nothing to do with the physical implementation of computers.
??? An algorithm that executes in O(n) time executes in O(n) whether the computer is silicon, DNA, mechanical, vacuum tubes, or whatever else you build a Turing machine in. I think "computer", to most people almost always refers to the physical implementation of a computer.
Once parallelism enters the picture, our analysis must change. We have to weigh communication costs as well as computation costs, and sometimes we actually want to pick an algorithm which would, in a sequential world, be significantly slower.
In short, what "computer" you use can change how your analysis should proceed.
therefore it should not be called science? Software development involves experimentation, testing, trial and error, measurement, standards, and it's much more rigorous than any science.
Computer Science is a ill-advised conflation of two different fields, of which one is largely an engineering discipline and one is essentially applied mathematics that the engineering side draws upon for tools and techniques. Neither is a "science" under any plausible definition of the word.
At least at my school (U of Washington), CSE and Informatics are completely different programs. Our Info department is basically information science, web design, and watered-down programming. They are two distinct disciplines and should not become one.
"Computer Science" encompasses fields that do depend on experimentation and are concerned with real implementations. There's more to Computer Science than theory.
Being concerned with real implementations largely falls under the category of engineering, and I have yet to hear a concrete example of depending on experimentation that doesn't amount to "we have to use the scientific method to make sense of shoddy, undocumented work by other programmers", which I have a hard time regarding as central to the field, even if regrettably necessary.
12 comments
[ 4.0 ms ] story [ 39.6 ms ] threadOnce parallelism enters the picture, our analysis must change. We have to weigh communication costs as well as computation costs, and sometimes we actually want to pick an algorithm which would, in a sequential world, be significantly slower.
In short, what "computer" you use can change how your analysis should proceed.
therefore it should not be called science? Software development involves experimentation, testing, trial and error, measurement, standards, and it's much more rigorous than any science.
Related: http://news.ycombinator.com/item?id=690798