Slashdot Mirror
Google Engineer Decries Complexity of Java, C++
snydeq writes "Google distinguished engineer Rob Pike ripped the use of Java and C++ during his keynote at OSCON, saying that these 'industrial programming languages' are way too complex and not adequately suited for today's computing environments. 'I think these languages are too hard to use, too subtle, too intricate. They're far too verbose and their subtlety, intricacy and verbosity seem to be increasing over time. They're oversold, and used far too broadly,' Pike said. 'How do we have stuff like this [get to be] the standard way of computing that is taught in schools and is used in industry? [This sort of programming] is very bureaucratic. Every step must be justified to the compiler.' Pike also spoke out against the performance of interpreted languages and dynamic typing."
Should have RTFA I guess, I now realize Mr Pike just talks in circles and really didn't have anything of value to say other than 'programming is hard'.
No, he doesn't. TFA-writer Joab Jackson talks in circles and doesn't have anything of value to say. Mr. Pike, on the other hand, appears to be saying that Google Go fixes a lot of unnecessary complexity in Java and C++.
His keynote isn't linked from either the Slashdot summary or TFA, but can be seen here: http://www.youtube.com/watch?v=5kj5ApnhPAE
Want a high quality FOSS RTS game? Try Warzone 2100!
I must say I rarely find a comment on /. that I agree as much as I do with yours.
C and Python march hand in hand, one is for machine performance, the other is for programmer performance. If someone thinks C is too complex or too hard to learn then he shouldn't be working with programming computers, he's likely to cause great damage sooner or later.
However, there's one point where C will need a new approach: multiprocessing is coming. Since it seems like Moore's law has hit the ceiling at 3 GHz CPU speeds, all progress in performance for the foreseeable future will come from increasing the number of CPUs and cores working together.
I have done a lot of programming in multithreads using the pthread library lately and I feel that something better is needed, pthread is not close enough to the metal. I think some new fundamental elements may be needed in the language.
C is so great for programming because it mirrors the hardware closely. For instance, pointers work so well because they represent memory addresses. Before I learned C I had worked with Fortran, I still have some programs I wrote over 25 years ago. Today I look at those old Fortran programs and I wonder why I did some things the way I did. I see some convoluted loops and wonder why I did that because, with a quarter century hindsight on using pointers, I create almost instinctively the most efficient set of pointers to handle a data structure.
What programmers often don't realize is that the correct data structure may get orders of magnitude improvement in performance. To give one example, years ago, when I studied artificial neural networks, I read an article in the Doctor Dobb's magazine (January 1989, page 32, "Neural Networks and Noise Filtering" by Casey Klimasaukas). It was a good article, but the source code in C that came with it sucked. There was a struct _pe representing a processing element and each struct _pe had an array of struct _conn representing the connections to that element.
The problem is that in an artificial neural network what each neuron is doing is, basically, a convolution of two arrays. To do that efficiently in hardware you need to have the array elements contiguous in memory. When you put the connection weight in a structure together with other data you will not have that value contiguous with the weights of the other connections.
From an "object oriented" point of view that program was perfect. But if you want to use your multi-core CPU with that, the program sucks. That's the benefit you can get from programming in C that you won't get with other languages.
And don't tell me that raw performance does not matter because you can always get faster hardware. CPU clock speed has stopped at 3 GHz, we must learn to use our multicores if we want to evolve from now on.