Re: Computer operating systems
On Apr 8, 1:53 am, stan <smo...@exis.net> wrote:
James Kanze wrote:
On Apr 7, 3:06 am, stan <smo...@exis.net> wrote:
For a little on topic content, is there any operting system
written in C++? I think parts of windows is C++ but is the
whole thing? It's hard to seperate library and system at
times, but here I'm wondering about the basic kernel in c++.
At the very lowest level of the kernel, it has to be assembler.
There's no way to write a context switch in C, let alone C++.
I was allowing for some assembly, I should hve been clearer.
I understood that, but when you speak of "the basic kernel", how
high do you go? A well designed OS will be extremely layered,
and the "basic kernel" won't necessarily be much more than the
scheduler lookup and context switch. (In the one OS I wrote,
about 10 machine instructions in all.)
With regards to C++, Symbian seems to be mostly C++.
Interesting, I've heard of it but I'm not very familiar with
I've never programmed to either, but from what I gather, it's
more wide-spread than Windows (since it is running on most of
your cellular phones).
There was also a distributed OS developed in France some 15 (or)
more years ago that was largely written in C++. Chorus, or
something like that. It was bought out by Sun, and presumably
parts of it have made their way into Solaris.
All the operating system and driver stuff I've ever been
involved with has been non object stuff and I was trying to
imagine what an object oriented approach might look like.
Interesting. Although it didn't go by that name, most OS's have
been OO long before OO became popular. Think about it: the
basic principle of OO is polymorphism, right? You call a
function, and it's actual implemnetation and behavior depend on
the type of object you call it on. In other words, when you
call write(), if it is OO, the actual behavior will depend on
whether the file descripter refers to a file, a console, a
socket, or... The kernels of most OS's go back to before C++,
but all of the ones I've worked on had a manually implemented
vtable for the I/O. I'd say that C++ would be a natural for
that part, at least.
When I wrote my OS, I had one system request, respond, which
depended on the type of object whose handle it got: if you
passed it a process, it would start it; a message queue, it
would send a message; etc. Originally, I implemented it as part
of the timer queue: when you made a timer request, you'd pass it
a handle and a "message", and what happened when the timer event
occurred depended on the type the handle referred to. I then
decided that it might be more generally interesting: you'd send
a request to another process, with a handle where you wanted the
response, and the other process didn't have to know how you
wanted the response sent; the OS took care of it all. This was
in 1979, and I'd never heard of OO or C++ at the time, so in
fact, I implemented the polymorphism by means of a switch, but
the idea was there.
James Kanze (GABI Software) email:firstname.lastname@example.org
Conseils en informatique orient=E9e objet/
Beratung in objektorientierter Datenverarbeitung
9 place S=E9mard, 78210 St.-Cyr-l'=C9cole, France, +33 (0)1 30 23 00 34