[3616] in java-interest
Re: Java is not slow ! (was: Introduction to Java - Questions)
daemon@ATHENA.MIT.EDU (Craig R. McClanahan)
Fri Nov 17 17:48:18 1995
Date: Fri, 17 Nov 1995 09:50:30 +0800
From: crm@perpetua.dat.com (Craig R. McClanahan)
To: stg!zog@uunet.uu.net
Cc: uunet!inria.fr!Anselm.BairdSmith@uunet.uu.net,
uunet!eshop.com!mlorton@uunet.uu.net,
uunet!xetron2.xetron.com!alfv@uunet.uu.net,
uunet!java.Eng.Sun.COM!java-interest@uunet.uu.net
In-Reply-To: <9511162335.AA20496@mcbain.stg.com> (stg!zog@uunet.uu.net)
>>>>> "Chris" == Chris Herzog <stg!zog@uunet.uu.net> writes:
[Earlier discussion elided ...]
Chris> At the risk of sounding trite, real users (whicha lot of us
Chris> ain't), don't care unless something makes their life
Chris> better. Often times something needs to be 2-4 times better
Chris> just to justify a change.
One of the problems, though, is that users don't always know what they
SHOULD care about. For example, implementation quality is important, but
is not always apparent or relevant.
To go along with the "3/8 inch drill" metaphor, last summer I was pulling
some nails out of an old deck, with a hammer -- and I broke the head off.
Why? Because the implementation of the cheapo hammer I bought wasn't up to
the task that I was using it for, even though the hammer's claws fit very
nicely over the nails I was pulling! Quality differences affect us,
even when we don't know what to look for (I'm a software jock -- hardware
is magic! :-).
Chris> Measurements like this against other things on a _relative_
Chris> basis is one thing - lot's of real world solutions are
Chris> often judged on absolute basis where as in the above
Chris> example, the technology is not the deal but the results
Chris> are...
I have no problem with measuring things on an absolute basis, if you're
measuring something that is relevant. Consider:
* Is it possible to implement something like Java in more than
one way? Sure -- we already have two major ones (Sun and
Netscape).
* Is the performance of these two implementations on a common
platform, running the same applets, identical? I haven't
measured it scientifically, but there sure appear to be
differences.
* If there ARE differences, then clearly something other than
the language itself is causing that difference. In the case
of Java, they're even running the same bytecodes!
* The only possible difference in such a case is the implementation
of the bytecode interpreter. From the user's perspective that
may seem to be part of the "language", but that clearly isn't the
case -- it's part of a particular vendor's implementation.
In the days before microcomputers (and even somewhat before minicomputers),
IBM got early on the bandwagon of structured programming by publishing a
compiler for PL/1 -- an early attempt at merging the features of COBOL and
Fortran into a language that would serve multiple needs. The problem was,
the compiler that was released was buggy, slow, and produced inefficient
code (relative to compilers for different languages that compiled more
efficient programs to execute the same algorithms). As a result, the PL/1
language itself got a reputation for being slow.
If this "slowness" were an attribute of the language itself, rather than
a particular implementation, we'd have no business worrying about benchmarks
like whetstones or whatever, that compare the execution speed of code from
different compilers, starting from identical sources, and executing on the
same runtime platform. (Of course, the availability of particular features
in a language can make efficient code easier or harder to produce, and that
IS a language issue, but it's not the only thing that affects performance).
Chris> --
Chris> Chris Herzog Software Technologies Group zog@stg.com - This
Chris> message was sent to the java-interest mailing list Info:
Chris> send 'help' to java-interest-request@java.sun.com
----------
Craig R. McClanahan EMAIL: crm@dat.com
DAT Services Phone: 503-526-6405
Beaverton, OR, USA Fax: 503-526-6442
-
This message was sent to the java-interest mailing list
Info: send 'help' to java-interest-request@java.sun.com
