[svlug] Re: Thread Programming and Java

[Karen Shaeffer]

On Thu, Jun 08, 2000 at 09:32:55AM -0400, Justin F. Knotzke wrote:
> On Wed, Jun 07, 2000 at 10:28:21AM -0700, Karen Shaeffer wrote:
> 
> > code that runs. Are you interested in working with this crap?
> 
> 	I work with this "crap". 

Let me apologize to anyone who is offended by my statement above. I really
mean no disrespect to anyone. I have views and a preference and have
expressed them. I have no problem with others seeing it differently.

> 	And why should I either? As for the tradeoffs there have been many
> 	studies (one published on slashdot) that shoes that Java is not as
> 	slow as many people think it is and in some instances its faster. 


If you trace through the thread leading up to my comments, you'll see that
the assertion was made advocating the *need for the Linux kernel to be
optimized to support Java*. Java does not utilize the hardware efficiently,
and the Linux kernel is designed for optimized execution at the hardware
level. You won't find any credible studies disputing my views here.

According to Operating System Theory: An OS is designed to optimize the
execution of the hardware--implementing mechanisms and not incorporating
any policy. Linux adopts this philosophy. In my opinion, languages
should be coded to expose the full power of the OS/Hardware engine they
run on. (Java fails this criteria, so I don't use it.) One must factor
in platform archtitectures, such as Linux clustering, to really see the
fully generalized equation.

> 
> 	There are many instances where C/C++ is better but there are many
> 	other instances where Java is superior. The right tool for the
> 	right problem.

I have no problem with this statement. Just don't come back to the kernel
folks and tell them to corrupt their system so you can expand the scope of
relevance of Java. Which is exactly what was advocated in this thread prior
to my comments. (And I realize you have not done that here.)

>       Not everyone in the
> 	programming world writes device drivers. There was a time where
> 	"real programmers" wrote in assembler and not in C.

Well, that is true as well. But these individuals should also refrain from
telling the programmers who live close to the metal how to optimize their
code. The programmers who live close to the metal should write code that is
*policy* free. To write code that skews performance to Java is to inject
*policy* into the code. This is not likely to happen in the Linux kernel.
And it shouldn't--because doing so would violate an axiomatic principle of
operating system theory.

I enjoyed your comments Justin. Thanks.

c,
-- 
 Karen Shaeffer
 Neuralscape; Santa Cruz, Ca. 95060
 shaeffer at neuralscape.com  http://www.neuralscape.com