Another aspect not always noticed is the speed of the compiler. Because Java compilers don't perform any compile-time optimizations, they are significantly faster than C++ compilers. This is very important when dealing with very large codebases.
Of course, Python doesn't get compiled at all, and I've heard it argued that interpreted languages will be used exclusively in the future with very large code bases for that very reason.
Regards,
Chris
Berin Loritsch wrote:
Stefano Mazzocchi wrote:
on 6/24/03 6:55 AM Dirk-Willem van Gulik wrote:
Then perhaps my observation means absolutely nothing - and I should really
try to get my mind around a fundamentally different development model (and
some aspect you call WORA).
Oh, sorry, WORA := Write Once Run Anywhere. It's java's first commandament. Basically, it's bullshit: java runs everywhere because all virtual machines descend from the same codebase (in fact, those exotic virtual machines like Kaffe or natively-gcc-compiled are not used because the number of small incompatibilities/deficiencies is simply too big).
I guess we aren't on the topic of ASF membershipe anymore ;P
WORA translates automatically into Java's biggest sin: native code. Java programmers were tough the religion of "java purity" as the only way to purify their souls from the "sin of native code".
This is the reason why we basically we have mod_* where * is any
programming language, but not mod_java, there is no java API that mimics
the HTTPD API because we preferred to avoid the "sin" of doing JNI (java
native interface) and preferred the socket disconnected way with mod_jserv.
To be honest, Stefano, I partially agree with you--though not completely.
To me one of Java's strengths is the ability to develop on Windows and deploy
on a real operating system (read UNIX). It's for that reason that I am
not too thrilled about .NET--although in my brief brush with C# and the suite
of open source tools supporting it there are many features to like (and some
to dislike).
I don't see JNI necessarily as a sin, but I do see it as a management problem.
Just because a library works flawlessly in Linux systems doesn't mean it will
do that for Windows based systems, and vice-versa. Typically the library will
be strongest on the platform that the developer of the library is strongest.
That can be a problem if there are key features that you need in a library,
but it is unstable on the platform you are deploying.
SEDA's nbio package is awesome--on UNIX. It is much more efficient than Java's
nio package.
Another major strength for Java is that there is one API for all platforms.
That means I can develop with a very high degree of confidence that what I
write will work on platforms I care about. It also makes me much more efficient
as I don't have to worry about some low level details or what the exact
differences between POSIX calls on Windows and true POSIX calls are.
There is alot of advantage that Java provides--but it is not a panacea.
note that for apache 1.3.x, JNI would have been hard because of the multi-process environment, but for apache 2.0, a JNI-based mod_java is perfectly valid architecturarely, but nobody works on it because of this "sin" syndrome.
:) I think it has more to do with all the choices they have available,
and it is not a particular itch that many people have to scratch. It takes
an inovator to proove that folks can't live without it. It's the "Build it
and they will come" syndrome. You always run the risk of nobody comming.
Also, java programmers tend *not* to have any knowledge of things like "how to link a library in a native environment". basically they are totally isolated, which leads to concepts such as server microkernel architectures (avalon Phoenix) which look cool from a purely architectural perspective, but are totally weak from a security and stability perspective, because they use one JVM for the entire thing, a very weak setup.
Oooh, those are fighting words ;P
Seriously, Phoenix and other microkernel based systems have their place. Trying to put a mail server and a web server in the same JVM may not be all that great an idea, but that doesn't negate the need for something like Phoenix.
Security models are something that NO ARCHITECTURE EVER can get totally right. It takes a dedicated crew to stay on top of changes and try to anticipate the ways that people will try to break your system. That said, there is a lot to do in any Java based system.
Pier is probably the only person I know who has great capacity on both sides of the fence and he tried to add unix deamon-like capabilities to java but crushed into several JCP walls where native stuff is still seen as a sin.
There are isolates comming which provides ways of managing child JVMs and virtual JVMs. Not here yet.
Note how Java failed on the client side because of how slow swing is. Eclipse introduces SWT, something that Sun really disliked because of considered again a "sin" to have something OS-specific.
Whatever. It is catching on and folks are using it. I'm contemplating that myself. Then again, Swing performance is much better than it used to be.
Again, the culture difference between java and python, for example, is that python has OS-specific features, java does not and will never.
java is based on a common denominator. Python is based on giving access to what you have.
I want a common interface to common things. If Python can give me a common API for controlling windows, et. al. then I can use the same script on both Windows and UNIX. If I have a separate API for each one, then I have to write some glue code, detect what platform I am on, etc. etc.
If you can give me that with C++, then I am all over it, and I will leave the Java fold (just kidding--maybe).
personally, I feel ashamed I was not able to see that this WORA concept is just bullshit and that I wasn't able to see how mentally limiting this "pure java" thing is.
But it's far from common to have language open mindness in the java world and this is due to this purity religion.
Every language has its zealots. Every platform has its zealots. Practical
people use what works. That's me. I have to work with both Windows and
Unix platforms all the time. I don't have to worry about embedded programing
or anything like that. I don't want to waste time when I bounce back and
forth. For now, I get that with Java. If the Mono project takes off, then
there is a good chance to use .NET based stuff (or at least C#) which has
the same pragmatic idea as Jython.
We already have a port of the Avalon framework in sandbox for .NET. I am playing with that (at work we will have to know .NET in about a year--I plan to be ready with tools that I like).
All that said, I use what works.
