I've had to do a lot of performance tuning, and if there
is anything I've found it is that with Java the specifics
matter a lot more than the generalities.
Five or six years ago that wouldn't have been as true,
but Hotspot and JVM internals have changed so much that
how you construct a performance comparison
test completely determines the results you get. If the test
doesn't match what you are doing, the conclusions are
100% useless. Not close. Not good enough most of the time.
Just useless. You really have to keep that in mind when
reading articles that try to compare other languages to Java.
The conclusions could be either completely accurate or
completely irrelevant to you, but most of the time they'll
be irrelevant. Doesn't mean they aren't interesting,
but they won't be much of a predictor for anything.
The articles may have one use though. I suspect that
any time an article comes out that spots a particular
feature of Java as being slow, a later Sun JVM
probably comes out that fixes the issue. Definitely
that happened with reflection and
string operations.
[ January 12, 2006: Message edited by: Reid M. Pinchback ]