The quality of the service, in the point of view of help to solve a problem we are encountering has significantly dropped. This is partly due to policy and partly due to the much lower quality and pertinence of the answers.
Contributors are there to earn points and just try whatever answer comes to their mind even if it's so vague or approximate that it is useless. Just trying, nothing to loose.
I guess the experts went away from it because the initial thrill has gone (sign of smartness) and they have better things to do then stare at a browser to catch new questions for no other rewards then getting points.
My impression is that stackoverflow was initially in a virtuous circle and now it's in a vicious circle.
I now have questions like how EXT4 files with holes impacts performance of an mmap ? Should I force a prior full fill of the file with dd before mmap ? I'll certainly get crap answers and I really don't know where to ask this to get a valuable answer. So I'll end up doing a benchmark, which has probably already been done many time or people knowing how EXT4 works could answer straight out of the book.
Asking a question typically leads to a handful of "shotgunned" answers. Yes, they do come immediately -- but the real story is, they are templated ("I'll fix it later, just lemme be the first to answer"), incomplete, and/or simply wrong.
A dozen explanatory comments later, the SO gun slingers finally understand the question and its implications. No, I really was NOT asking that. Neither that. Yes, that condition I mentioned in the question IS important and you cannot ignore it in your answer. You have an error there. That won't work.
Happened to me. And that person responding completely aside of the question voted to get my question closed and erased its answer. I was asking for axisting asynchronous signal slot library to avoid implementing mine. Another guy proposed to use QT and another one to use boost's signal which is not asynchronous. The later guy commented to the QT guy that boost was so much lighter and faster. Boost signal is known to be the less performant, but because its boost it must be the best. This guy also voted to close my question once I said I don't want to depend on boost. The question was finally closed and I implemented my self the signal slot lib with a perfect match to my requirements.
Contributors are there to earn points and just try whatever answer comes to their mind even if it's so vague or approximate that it is useless. Just trying, nothing to loose.
I guess the experts went away from it because the initial thrill has gone (sign of smartness) and they have better things to do then stare at a browser to catch new questions for no other rewards then getting points.
My impression is that stackoverflow was initially in a virtuous circle and now it's in a vicious circle.
I now have questions like how EXT4 files with holes impacts performance of an mmap ? Should I force a prior full fill of the file with dd before mmap ? I'll certainly get crap answers and I really don't know where to ask this to get a valuable answer. So I'll end up doing a benchmark, which has probably already been done many time or people knowing how EXT4 works could answer straight out of the book.