[ColorForth] OS is Not a Dirty Word

["Kurt B. Kaiser" <kbk@xxxxxxxxx>]

Jack Johnson wrote:
> http://www.pdos.lcs.mit.edu/exo/
> 
> The papers are worth a read.  They go back to this idea in an interesting
> way.  They looked at Unix and having a multiuser system needing to have
> simultaneous access to resources and they reengineered the way the OS
> deals with the essential problem at a very low level, so now the "OS" is
> essentially a set of libraries that ride on top of the kernel.  The fun
> part is that because we really are just dealing with the hardware, the
> abstraction is arbitrary, so they did things like implement a Web server
> with it's own TCP/IP stack and storage system to squeeze out the
> performance (which worked), without mucking about with the same subsystems
> that handle the same tasks for everyday users.
> 
> Worth a read.

Yes, this is interesting; I had not seen it before. Thanks.

They lower the app/OS boundary and then allow the app to inject code into the
protected space of the kernel. To permit that safely, they use 1. Automated
code inspections 2. Inlined cross-domain calls, and 3. Type safe languages.

Forth is not a type safe language! 

The reasons for doing it are to allow the app to correct for inefficient, slow
implementations in the kernel (well, fix it!) and to allow for implementations
which emphasize different ways of accessing resources.

In the absence of hardware protection, which as far as I know doesn't exist in
Chuck's designs, these issues are rather moot. I stand by my point of uniform
resource access, but I guess the lesson here is keep it as low level as
possible.

Regards, KBK
-- 
   K u r t    B.   K a i s e r
   k b k @@ s h o r e .. n e t
------------------------------
------------------------

To Unsubscribe from this list, send mail to Mdaemon@xxxxxxxxxxxxxxxxxx with:
unsubscribe ColorForth
as the first and only line within the message body
Problems   -   List-Admin@xxxxxxxxxxxxxxxxxx
Main ColorForth site   -   http://www.colorforth.com