Head to Head

[kmj]

Okay,

I'm really interested in seeing some real world comparisons between java and .NET.. For any number of people who say Java's fast enough, there are just as many who say it's too slow.

This is what I'm thinking: we come up with a (few?) apps to benchmark.

A .NET buff writes a version with .NET (any language), and gives detailed instructions on how to install .NET and run the application.

A Java buff does the same with a java1.4 application.

An "independent third party" writes the app in native code, probably C++ or C.

People download the apps, and run them, and reply with their findings... then we try to come together with a consensus.

Now, the main purpose of this is to ensure anyone can see first hand (meaning, running the application) how they compare; not just reading some webpage results. This also means, since the code will be available, that people will be able to check to make sure a programmer isn't doing anything silly to slow down his code unnecessarily.

What is needed:

1) ideas for applications; basically specifications just like for most of the other competitions. In fact, a number of previous competitions would probably work fine as apps.

2) A willing Java Coder

3) A willing NT Coder

4 ) A willing "native" coder


What do you guys thing about this?

[jemfinch]

As long as the application is non-trivial and different programmers are programming it, it's not really a competition between the languages. The space of possible algorithms and data structures possible for any given non-trivial program is just too big to expect that three implementations written by three different programmers are going to use the same algorithms and data structures.

Jeremy

[gish]

i'm in ...now what should we develop???

G

[kmj]

jemfinch: well, we can specify which algorithm is to be used.. and keep the programs relatively small, but just execute the same code a number of times.

[Strike]

The only problem I can see with this is that Microsoft is really the only platform .NET is ready to run on. You _know_ Microsoft hates Java and does basically everything it can to make it look bad. On the other hand, it is willing to do whatever it can to make .NET look good.

In other words, it'd be like running a speed test between IE and Mozilla on Windows. IE is integrated into the OS, so is .NET.

[gish]

I see what you are getting at....I beleive this to affect the results as well.......I never really though of that running a java app on my XP box....wow...maybe this is why they are slow....any ways......is that opensource .net framework deady for linux yet...sorry I forget the name...??....

[Dru Lee Parsec]

But we don't have to run Java as an applet (which uses the browser) run it as an application and simply install the Sun JRE. That should eliminate any anti-Java bias code within IE.

Now, does the code for any of the MS OS's have anti-java code in it? I don't know but I wouldn't put it past them. Back in the Dos 3.0 days the mantra at MS was "Dos isn't done until Lotus won't run!"

Also, a particular program I wrote years ago ran as much as 3 times faster on the same physical computer when it was booted to Linux rather than Windows NT 4. So who knows if MS is trying to trash Java in it's OS code. Maybe.

But even if they are it's just something we have to deal with. Maybe we should be flattered that Java was something important enough for MS to attempt to "Embrace, Extend, and Extinguish".

[GnuVince]

I think I recall Dru saying that his first contest ran faster on his Linux box. SO, yes Java is possibly faster on Linux

Lunix si teh rockz@!!!~

[gish]

I do know that MS does not support Sun's Java officially anymore.....I went to install Sun's java on my XP box, from MS site, and got an error message that it is no longer supported by them!

G