Resent-Date: Wed, 6 Dec 1995 17:42:42 -0500
From: Dave Dittrich <dittrich@cac.washington.edu>
Message-Id: <9512062242.AA01340@red3.cac.washington.edu>
Subject: Re: Sound in pre2 works (and other status) (fwd)
To: dunham@cl-next4.cl.msu.edu
Date: Wed, 6 Dec 1995 14:42:02 -0800 (PST)
Cc: java-linux@java.blackdown.org
In-Reply-To: <ML-1.3.2.818251382.1541.dunham@notung.msu.edu> from "Steve Dunham" at Dec 6, 95 07:03:02 am
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length:       4128
Resent-Message-Id: <"GZKEw1.0.OA4.kmXnm"@liquor>
Resent-From: java-linux@java.blackdown.org
Precedence: list
Resent-Sender: java-linux-request@java.blackdown.org

My two cents worth.

> >    I'm already running 1.3 with no problems at all. I vote go for it. 

> >  Never fear to live on the bleeding edge!

My guess is that you have never supported multiple users/programs and
had to deal with the breaking one thing to "fix" another.

> > Seriously, in my experience 
> >  1.3.45 has been very stable.

Hmm.  You installed a development OS and haven't had a crash in a few
days.  Does this mean that it will still be up in 120 days?  "Stable"
is more than not having problems when exercising a limited subset of
an operating system's functionality.  It also has to do with frequency
and breadth of changes introduced in incremental updates.  I wouldn't
call something that is in its 45 revision in just a few months "stable".

> Also: 
> 
>   It takes some hackery to get 1.2.x to compile ELF anyways

I have Caldera Preview II (RedHat 2.0) installed on my system (which
has never crashed on me; EVER) and didn't need to do any "hackery" at
all.  Perhaps the problem is your choice of distribution and having to
incrementally upgrade bits/pieces of things.  If this is to save
money, you've just discovered why "free" is a misleading term.

>   There are some very stable 1.3.x kernels.

This sounds to me like you mean "you can pick one out of .45
possibilities for a kernel to base development of Java on and it may
not crash on you."

>   The later 1.3.x kernels have better scheduling and swapping.

This may be true.  I haven't wished to cut myself on the bleeding edge
yet to find out.

>   A lot of people are forced to used 1.3.x because they have devices
>    (CD-ROMS - I believe) that only work in 1.3.x

Not to be argumentative, but this is conjecture.  Percentages?  Proof?

>   The PPP in 1.3.x is better.

They both work.  A small improvement in one area may not outweigh an
unstable (but critical) feature elsewhere (known at this point or
not).

>   1.3.x is more modular.

This could also be true.

>   You should `Do the right thing'(TM) - I don't have in depth knowledge of
>   the problems involved here, but it seems that kernel threads are `The
>   Right Thing'(TM).

Thread support is desireable, but not critical at this point in time
(my opinion).

I'd personally rather minimize the variables that Randy has
to deal with so that he can concentrate on debugging a working (get it
working first; optimize later) Java development environment.  In my
experience, if you use too many development (read "variable")
components to solve a problem, you spend far more overhead time
debugging and fixing the variables than you do on the end result
(learning/developing Java applications).

My vote is that Randy stick with the most common 1.2.x ELF kernel that
is used (seems to me to be something akin to RedHat 2.0) to minimize
the impact of kernel and development tool (e.g., compiler, linker,
etc.)  instability and get the JDK itself running and stable (and,
perhaps, integrated into the Sun source base so that Linux is more
respected as a development platform by Sun et al).  Then worry about
getting it going with the newer 1.3.x kernels next year some time.

I'll just end by saying that (in my opinion) the only way Java is
going to get the competitive lead on Microsoft that it needs is for
the development tools to be working ASAP and for LOTS OF PEOPLE to
start using it. Let's look at this with a professional software
development slant for a minute.  It seems extremely unwise to waste
time trying to use the latest, "greatest" (where "greatest" means
hacker's wet dream, rather than business man's trusted tool)
development kernel to do further development/debugging on an extremely
high demand development tool.  You're not going to see Bill Gates
announcing his intention to use Windows97 or NT 5.0 for Blackbird in
his press conference tomorrow; he'll use "stable" NT 3.5 and the rest
of last decade's operating system technology to build tomorrow's Web
foundation.

My words, my opnions, blah, blah, blah...

-- 
Dave Dittrich
dittrich@cac.washington.edu

<a href="http://www.washington.edu/People/dad/">
Dave Dittrich / dittrich@cac.washington.edu</a>