[unios] Project update.

Pat Wendorf beholder@ican.net
Fri, 05 Mar 1999 18:58:32 -0500


Pieter Dumon wrote:

> >
> > I believe Beholder will be recieving this from the Unios at tunes.org list, The
> > message is primarily to him but it concerns everybody. After much concideration
> > I have decided to change tack on the entire project. The orrigional goal was a
> > totally hardware independant software platform. I nolonger believe that this is
> > fesable in the near term.
>
> It is not even necessary to make a complete hardware-independent software
> platform. You can hide most hardware-specific things from the user, but at
> least the supervisor must be able to do hardware settings and OS
> optimizations. Also, it must be possible for programs to detect the number
> of cpu's in the system etc. Like that, programs can easily define how to
> divide the workload. The OS alone can't do that. Take an example to the
> rc5 contest : you can run the client numbercrunchers with options on how
> many cpus your computer has etc. Very effective. You don't need to hide
> all hardware from the user - as long as it is safe.

Agreed.  The hardware abstract model I envision does quite the opposite, it allows the
supervisor to modify any, and all hardware settings.  The whole point is to make the
hardware easy to program, and standardized methods across all platforms, not reduce the
ability to play with hardware settings.

> Also, it is not possible to make one environment that fits all needs. It's
> just _impossible_. Graphical workstations need a completely other
> environment than  webservers. But lots of software layers can be the same,
> off course, and it can almost all be programmed hardware-independent.

Agreed also.  With hardware abstraction, programs could use any I/O interface that makes
sense for situation.  I don't think anything should be forced on the programmer (like a
GUI).

> >       Where do we go from here? I believe that the best course of action at this time
> > is to take Linux or OSKit and customize it so that it meets my definition of
> > user friendly.
>
> Taking the Linux kernel or the OSKit has nothing to do about making
> something userfriendly, has it? These are low-level interfaces and only
> need to be understandable by OS programmers.
>
>
> > To that end I will spend the next 6 months finally learning LiNuX
> >  as it is today so I can start plan how to rewrite it.
>
> Is rewriting Linux just a good option? Linux takes a monolthic kernel
> aproach. It is modular, so it is very flexible, but far to difficult to
> get stable, due to the design. It is possible to make much more flexible
> models. You could base that work on the Linux kernel, but generally,
> you'll end up with more trouble than just to start from scratch (from
> scratch means usign the OSKit , for instance)
>
> > I will also start studying math in an attempt to come up with that new
> > discovery! :)
>
> :) Good luck.
> Perhaps you'll end up with a completely new computing model. An
> alternative to the Turing model or so...

I can't imagine one :)

--
-----------------------------
Pat Wendorf
UniOS Group
http://members.xoom.com/unios
beholder@ican.net
ICQ: 1503733
-----------------------------