[cap-talk] Dan Bernstein's qmail security lessons paper

[Ivan Krstić]

On Dec 17, 2007, at 4:05 PM, James A. Donald wrote:
> The test you cite fails to measure the bad Java startup,

It doesn't; it's listed as the 'startup' test, where server-mode  
Hotspot is 92 times slower to print 'hello world' than C. (It would be  
faster in client mode.)

> The Java programs usually took up twenty to thirty times
> as much memory as C++.

The linked test agrees with this; Hotspot used between 1.4x to 29x  
times more memory than C, mean of 16.27x.

> my subjective experience: that Java really sucks.

My subjective experience agrees entirely, but the pandemic of  
performance hand-waving in our industry still grates on me. Anyone's  
free to say "I dislike x because of y". When they say "x sucks because  
it's slow" (especially if they omit the "when doing y" fragment!), I  
want to see numbers, and they better understand basic statistics and  
experiment design. Then there's also the stupidity of how  
"performance" has come to mean "as fast as possible" for no good  
reason[0]. Anyway, this is off-topic for the list, so I'll leave it  
here. Cheers,


[0] http://zestyping.livejournal.com/193260.html

--
Ivan Krstić <krstic at solarsail.hcs.harvard.edu> | http://radian.org