Re: [jos-general] Re: Multiple processes

[John L. <je...@pi...>]

> > > 	Right.  That is, you're saying (and I'm agreeing that) multiple
> > > Java processes across multiple native processes is the easiest to get
> > > running (it's already here) but the hardest to adapt to do
> > > (object-oriented) IPC and memory conservation (etc)?
> > 
> > Yes.
> 
> 	OK, now I'm clear.  (It's neat that we can have a little
> discussion like this, but I digress... my 'ooh, cool toys!' geek reaction
> is getting a little out of hand.)

:-) It's quite welcome.
 
> > the source to the class libraries and there isn't that much native code.
> 
> 	Well, maintaining a set of patches to fix whatever is wrong with
> them will be amazingly annoying.  Oh well :)

Yes it will be, but the code will have to be patched anyway, eg accessing
files in the linux kernel is different to userspace, same for
networking. And that's the majority of the code.

> > An issue however, is whether we'll support JNI for apps running on our
> > OS. There might be a problem there. 
> 
> 	I would hope not, philosophically, but since you're looking at
> doing this on Linux, many of the problems I would have had doing it on
> jJOS vanish.  Primarily, I mentioned JNI because it's how the class
> libraries are written, and I'd like to adopt one more-or-less intact, to
> save work...

JNI in kissme/teaseme is implemented fully (as far as I'm aware, I didn't
write the code and I haven't done any kind of comprehensive test).

The current classpath libraries work fine with the current JNI
implementation (and so far work fine in the linux kernel). 

John Leuner