Re: Ada Stuff and some confusion

["Marin David Condic" <marin.condic.auntie.spam@pacemicro.com>]

"chris.danx" <chris.danx@ntlworld.com> wrote in message
news:Pb6z6.5897$%W5.648159@news6-win.server.ntlworld.com...
> > But they seem to be shooting for something more along the
> > lines of a WinNT replacement & that's not my interest.
>
> I don't think NT is a good system, it's certainly better than 9X but
still.
> I also don't rate Unix/Linux much either, but this is cos' it's old
> (architecture-wise).  NT is a hybrid microkernel and has some nice
features
> but it's still has many faults and shouldn't be modelled by anyone.
>
Said it was a *replacement* for WinNT - not a WinNT clone. By this, I meant
it is intended to be an OS that would drive around a PC/Workstation -
standalone or possibly in a network. There are other kinds of OS's with
different objectives. From what I can see of AdaOS, there is no intention to
build a kernel that would be acceptable for embedded/realtime work - hence
it is of only peripherial interest to me.


> You said you wanted a "a solid realtime, embedded subset kernel", why
don't
> you consult with Hermann and me, and we'll see what comes from it.  This
is
> just the beginning and we haven't finalised design and any such like.  You
> could make some suggestions, and if you're interested in contributing code
> then I can talk to Hermann and see what he thinks.  We're doing this in
our
> spare time, and we understand things like work and life have to go on and
> take priority.  We're not going to pressure you.  You could become a
member.
> You might just contribute to the project that could make windows look like
> the Dinosaur of the OS world.
>
Well, I'm not into full-blown OS design and any "theory" that I once knew
for OS design has probably been vastly expanded upon since when I was in
school. The basic requirement for an RTOS is predictability. You absolutely
have to have guaranteed latency for things like interrupt handling, context
switching, etc. You also have to guarantee priorities so that the RT
programmer can decide what is most important and know it won't get
sidelined. (Don't stop my #1 priority process just because the OS thinks it
has something more important to do! :-) Beyond that, you've got to provide
means for getting at a variety of hardware either directly or through
drivers (again, with predictable latency & priority!) and you need a few
"creature comforts" like communications mechanisms, etc. The rest is just
providing libraries of subprograms that get specialized jobs done for you.

The problem with an RTOS vs a Workstation OS is that they aim at different
goals and sometimes there is a conflict. If it is possible for a programmer
to write a program that doesn't give up the processor - even to the OS -
this means it is possible to write ill-behaved software. An RTOS has to let
you do that because the application has things to do that can't be
interrupted or delayed. Typically, the RTOS runs only one or a small handful
of very specialized apps and you can get your fingers around the neck of the
guy who does anything stupid or malicious, so this is OK. For a
general-purpose workstation OS, such a capability is rather dangerous and
could lead to all sorts of security holes. (Some workstation RTOSs only
allow certain things to happen if you are "privileged" to do so, but making
sure you don't have security holes can be much more difficult than simply
saying "You can't do that!" and being done with it.)

At the present time, between my "real" job and a small handful of other
sideline projects, I simply don't have time to go tilting at OS windmills.
However, if you guys have questions I might answer, feel free to drop me a
line. As I said, if I could find the RTOS kernel I wanted, there's an app
that is screaming for a superior RTOS & it would be an excellent opportunity
for Ada to break into something new & establish itself. I'll keep your
website bookmarked & check on it periodically. Thanks.

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/