[Prev][Next][Index][Thread]

Re: OSkit & languages that don't suck.



> It would REALLY make my life easy if I could work on the object level
> and focus on designs to work rather than getting code to work...

Sorry, you can only get away with that if you've got real programmers
working for you.  *Someone* has to do the real work of writing the code.

I wrote this to the list a few months ago, after answering some inquiry:
   p.s. If you don't have some experience with tweaking [the internals of]
   Linux or BSD or some other low-level system, the OSKit is probably
   not for you.  The OSKit makes building OS's hugely easier, but it's
   still *hard*, in an absolute sense.

---------------

That said, for the wider audience I'll mention that we have a fairly
long series of papers covering certain aspects of designing operating
systems tailored to run programs written in high-level languages,
starting with a bit in the SOSP'97 OSKit paper, but focusing on Java.
We also have software releases that provide an environmements for
safely running Java on the bare hardware: http://www.cs.utah.edu/flux/janos/

Languages for which we or our co-authors have experimented with
developing custom OSKit-based OS's, in chronological order:
        Java  ML  Smalltalk  Scheme

We have something like 4 years of R&D into the first one, Java,
although most of the lessons apply to all typesafe languages.  We have
a representative on the new Sun expert group that is refining a
related extension to the Java spec (JSR-121) and yesterday our student
Godmar Back defended his fine dissertation on the issues. (hurray!)
The Smalltalk deal was a one-shot demo.  The Scheme effort continues
here in related directions.  ML was done by colleagues at MIT.