| From: | Andy Wanless |
| Date: | 24 Jul 2000 at 21:22:23 |
| Subject: | Re: Executive Update |
On Mon, 24 Jul 2000, Matt Sealey wrote:
>
> > Didn't impress? Around 15-50 times faster? Pretty impressive, isn't it?
>
> No, as you explain below.
Hang on, I explain why it's impressive, and then you use that to justify
your claim that it isn't? Interesting ;)
> > Ok, so that's compared to Netscape on Linux. Compare it to Netscape on
> > Win98, and it's not quite as impressive, and hardly relevant anyway. If
> > the SDK ran on Windows, then there'd be something to compare it to fairly
> > there.
>
> Java is cross platform. 22x faster than Java should mean 22x faster than
> Java as a whole, not Java on Netscape for Linux, the slowest Java implementation,
For a start, nobody actually claimed 22x faster, it was 22%, wasn't it?
Secondly, Java isn't as cross platform as Sun hoped for.
Thirdly, why should it be uniformly faster than _every_ JVM? Netscape's is
slow, yes. Others all run code at different speed. So should someone
writing a JVM include some sort of magic code that detects when it's being
used for benchmarks and what it's being compared to and slow down if
necessary? ;)
Or maybe it should be faster by that exact same amount if you compare it
on 2 machines with the only difference being one has a processor twice as
fast?
> exclusively. As it happens, if you're used to the dog slow Java of Windows, then
> you're in for a treat when you use the SDK, 'cos it's SLOWER.
>
> Unless you deal entirely in string manipulation (unlikely)...
Or graphics? The main problem with all the versions of Java on Windows,
in my experience, is that it takes ages to start up. The main problem with
Java in general has been that doing any graphics, particularly any UI
stuff is stupidly slow and unresponsive at times. (Althought improving).
Of course, you were probably convinced it was going to be crap before you
saw those benchmarks ;) I wasn't convinced about their claims either, and
some of those numbers suggest that it's actually pretty impressive.
But then again, benchmarks like that are a pretty useless thing anyway. A
real test would be better. Some real application or something.
------------------------------------------------------------------------
The Most Expensive Thing You Can Haul Is An Empty Trailer.
http://click.egroups.com/1/5776/4/_/468125/_/964470167/
------------------------------------------------------------------------
Quote carefully and read all ADMIN:README mails