What a confused question. Responses inline.

On Sun, 3 Feb 2008 04:03:11 -0800 "Max Power" <mikehack@washington.edu> wrote:
> Has any form of UNIX (ignoring POSTIX compliant OS's that have an
> internal structure unrelated to Linix / BSD / Mach / ... like OS2 /
> etc..)

huh? First of all a note -- by invoking UNIX you are referring to
only those unices that are certified to carry the UNIX brand. This
goes beyond POSTIX (sic) compliance. So your parenthetical comment,
which as best I can make out says to ignore OS/2 for some reason,
even though it may be POSIX compliant, is curious since, e.g. Mach,
which you seem to want to include, is wildly different than the other
OSes you seem to want to include. By asking about UNIX you are already
excluding OS/2 as well as Linux and Mach.

But I guess that's not really what you mean by UNIX. And why would you
want to include Linux but not OS/2?

> ever had any kind of registry (for the core OS, not the
> applications; system daemons may be tracked by such a structure -- but
> not user daemons)?

huh? what is a system daemon and what is a user daemon?

> The Apple OSX (but not really AIX)

huh? What does AIX have to do with OSX?

> and RTOS variant QNX, as well as
> BEos are about the only substantial deviations from Unix in the modern
> era.

Actually OSX *is* UNIX, i.e. not only is it POSIX but it carries the
UNIX brand, and therefore by definition it is not a deviation,
substantial or otherwise, from Unix.

> Not a one has any kind of registry data structure that I am aware of.

AIX has a registry, as someone else replied. Solaris 10 has a
registry (SMF), although it's not consistently applied across the OS.

> With respect to modern Plug and Play issues, and some aspects of
> booting up -- a very limited registry would have some utility.

How so? Please elaborate. How would a registry possibly be useful
for PnP?

> There
> may even be some IEEE or ACM papers on variants of Unix that have
> tried this. I am not suggesting the concept be applied to Unix (non
> Intel) workstations or minicomputers.

Why would said registry, that you are suggesting might be useful, not
be useful for non-Intel computers? UNIX, and even unix, is not tied
to any specific hardware. Maybe what you are really looking for is
a BIOS alternative?

> I am not suggesting any design like anything in the Microsoft mold,
> just something to make storing and recovering (if necessary) core OS
> states more robust.

More robust? More robust than what?

Registries don't seem to provide robustness as far as I can see, in
fact quite the opposite. What they do provide is a consistent and
OS-provided, and BTW non-portable, interface for storing and
retrieving configuration information. That could be an advantage if
done right, but hardly applies to the "core OS". I would think that
the "core OS" could care less about a registry. It will have a
consistent interface for configuration regardless of what applications
use.

Most unix OS's do have what could be considered a registry for the
"core os", e.g. /etc/system on Solaris or /etc/sysctl.conf on many
others. It's very difficult to tell if that's what you're looking
for though.

-frank