Marty Leisner wrote:
> John Osterhout gave a good presentation at Usenix several years ago
> about "why threads aren't always good" (or something like that).
> You can find it on his web site (I think). It agrees with my thinking.
>
> It seems much of the argument of "lets use C++" is the same as
> "lets use threads". Sometimes it is the right answer, very often
> it is the wrong answer.
>
> One of the problems with threads, if it shared data gets corrupted,
> it is very, very difficult to analyze and debug...and it appears
> to be random in many cases...
>
> I've spent enough time debugging other peoples threads problems, the first
> question I ask is "why are you using threads?" Very seldom do I get a
> good answer to this question.
>
> Marty Leisner
The John Osterhout presentation is a very sane and balanced assessment of the
value of select based programming versus threads - recommended reading. I had
a similar presentation of my own once, but his is better so I threw mine
away. The only issue he doesn't adequately address is the scalability
limitation of select in multi-CPU environments. I guess it was written some
time ago, when multi-CPU systems were less significant. Perhaps he should
update it to give a properly balanced view for 1999's systems.
Steve
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/