Mark S. Miller
Tue, 26 Oct 1999 17:58:07 -0700
At 02:00 PM 10/26/99 , Paul Snively wrote:
> > >>... Unfortunately, the Cygnus Java is not far enough along for
> > >> them to test it.
> > >
> > >I'm not interested in Java compilers generating native code anyway.
> > Why not? I am.
>Eh... guess I'm hung up on the idea of being able to haul a little .jar file
[-] There is no conflict between compilation to native and portability
As your text below accurately explains:
>mean that someone couldn't/shouldn't run their .jar file through a compiler
>that goes from Java bytecodes to native code if they want to target a
>specific platform, however.
[+] Cygnus's ClassLoader should do this transparently.
E is written in pure Java and distributed as pure jvm bytecodes in a .jar file. E shouldn't care when or how a jvm implementation arranges to be able to execute the computation described by these bytecodes, as long as the jvm does so correctly.
I would like to distribute E as a fully linked .exe for various platforms *as well*, as that will allow installation without a prior jvm install, and because it should give astonishingly faster startup times than Java apps currently experience. However, we won't drop our pure-Java .jar distribution.