UNIX KERNEL SOURCE CODE ??

[Guy Helmer]

In <4526 at altos86.Altos.COM> ti at altos86.Altos.COM (Ti Kan) writes:

>In article <1991Jan3.112435.5509 at ux1.cso.uiuc.edu> bc at uxa.cso.uiuc.edu (Ben Cox) writes:
>>Minix is terrific.  [...]

>Minix is not at all like real UNIX.  The design and source is _completely_
>different.

What is real UNIX anymore?  Is it USL's UNIX?  BSD 4.3?  Mach with BSD
functionality?  SUN OS?  OSF/1?  XENIX?  BSD 4.4?
Each one is completely different from any other from the perspective
of the kernel programmer.

>            Minix supports no virtual memory, and many customary UNIX
>kernel functionality (such as the filesystem and process scheduling) are
>implemented as user processes that communicate with the kernel via a
>special message scheme.

I don't know that AT&T UNIX uses virtual memory that well; I seem
to recall that VM was kind of hacked in.  People are working
on adding VM to MINIX in some neat ways, but no one has posted
their code to the net yet.  The non-monlithic structure of MINIX
makes it easier to port to other machines than any other UNIX
kernel that originated from an AT&T source release.

As for message passing: I wonder what Mach does? :-)

>                         Minix is useful for a student learning about
>a possible way to approach operating system design, but it is _not_ useful
>as a tool to learn about the internals of real UNIX (whichever flavor of
>UNIX you prefer, be it System V, BSD, Xenix, or whatever) because of
>the drastic differences, and it is certainly _not_ a useful OS to do real
>work on due to its limitations.

Sure, it's limited the way UNIX Version 7 was limited (on PC's with 8086
or 80286 processors)---64K I&D.  The 80386 version of MINIX isn't
a production version yet, but it offers access to the full memory space
with 32 bit pointers, and GCC even runs on it.  Versions of MINIX for several
68000 processors exist, as well as one Sparc implementation.

MINIX is only four years old, but has gone through versions 1.1-1.2-1.3-1.5
with a 1.6 in beta and 2.0 being worked on.  Each version has contained
a huge increase in functionality, and 2.0 will conform to the POSIX spec.

At $179 per copy for MINIX (including full source & documentation),
it's quite a steal compared to the thousands of dollars per license
required for an AT&T source license for a university, or the 
hundreds of thousands for a company.  The binary license for AT&T SVR3
per copy costs probably almost as much as MINIX.

My point is that AT&T UNIX System V Release 4 isn't the
only UNIX kernel around, and in the future the major design differences
between kernels will become even less important than they are now.

As for MINIX being great or poor, it depends on what you want out of
it.  I say it is great, but that's because it gives me a UNIX base
that I can hack on and port stuff to.

>-Ti
>-- 
>Ti Kan | vorsprung durch technik!                                       \\\
>Internet: ti at altos.com                                                   \\\
>UUCP: ...!{sun|sco|pyramid|amdahl|uunet}!altos!ti                     /// \\\
>The opinions herein are not necessarily those of Altos.              ////////\
-- 
Guy Helmer                           helmer at sdnet.bitnet, uunet!dsuvax!ghelmer
work: DSU Computing Services, Business & Education Institute    (605) 256-5315
play: MidIX System Support Services                             (605) 256-2788
postnews: message content ambiguous; spurious information added as required