Richard Cownie (tich@pobox.com) on 8/27/09 wrote:
---------------------------
>JasonB (no@spam.com) on 8/26/09 wrote:
>---------------------------
>
>>In essence, he has based comments like "I suppose occasionally people manage to
>>write correct threaded code, but mostly it has lurking bugs" on his personal experience
>>and I've been arguing that his personal experience is insufficiently representative to make statements like that.
>
>Not just my personal experience, but a certain amount
>of reading. Here's the link to the "Problem With Threads"
>paper that I mentioned:
>
>http://www.eecs.berkeley.edu/Pubs/TechRpts/2006/EECS-2006-1.pdf
>
>One part describes a project done under exceptionally
>favorable circumstances:
>- designed from scratch for threading
>- careful design by EE/CS academics
>- coding by people of above-average competence
>- a rigorous process of code reviews and regression tests
>
>After the program had been in use for about 2 years, it
>hung up with a deadlock bug that had eluded all the
>careful design, review, and testing.
>
>I think we've had an interesting debate: evidently
>threading is working well for your project. And I think
>we've illuminated some of the factors that contribute
>to that result. Anyone considering parallelization can
>profit from reading this thread and considering how the
>pros and cons relate to their own circumstances -
>which very much have to include social/cultural factors
>as well as purely technical issues.
>
>Now I should quit posting for a while and go write
>some code :-)

Lots of the discussion here seems to not be about language at all. Richard is dealing with an environment with legacy code and not-particularly-great programmers. Jason is dealing with good programmers and code that has been written carefully from the beginning.

We could say that it's one of these "horses for courses" things? But, just as with horses there are general benefits of taking one course over another. The problem though, is how do you tell what those are?