Re: Linux isn't an operating system

Speed Racer (
Fri, 8 Mar 1996 01:55:20 -0500 (EST)

On Wed, 6 Mar 1996, Richard Stallman wrote:

> I learned to distinguish between an operating system and its kernel
> while working at MIT, since before starting the GNU project. That is
> why the GNU system is not the same as the GNU Hurd plus Mach.
> I was surprised to learn that some people consider "operating system"
> to be synonymous with "kernel". But they have quoted textbooks, so
> this usage seems to be well established.

I am surprised by this too. A number of people on this list, all of whom
seem very well-educated, point to numerous academic textbooks (I've counted
at least 3 so far), all of which describe an "operating system" as what we
would generally call a "kernel". However, I think that all of us have at
least passing familiarity with at least one other "operating system"; seeing
as the PC and Macintosh are far and away the most popular (at least it
appears that way to me), let's look at the "operating systems" for each of

PC's can use DOS/Win3.1, Win95 or WinNT. I've heard many people state
that the first pair isn't an operating system; this seems to be based
more in opinion than in any analysis of technical flaws. All three are
operating systems, all of which include a kernel as well as some
basic utilities (file management tools, disk management tools, a command
line interpreter & a graphical front end to all of the above). There are
a number of programs that are technically not essential to the operation
of the computer, but you would be hard pressed to get any real work done
without some basic ones. Macs have a similar configuration, except that
there's no CLI and the GUI is built into the kernel (forgive me for the
simplistic treatment). They too include some standard utilities. When I
talk installing the Windows 95 operating system, I don't mean that I've
put just the kernel on my machine; I mean that I've installed it and its
ancillary management tools.

> The other usage, which distinguishes the two terms, is also well
> established. Here's what the Feb 1966 issue of Linux Journal says, on
> page 7, in answer to the question "What is Linux?"
> Linux itself is the kernel, the "core" of the operating system,
> Most people use "Linux" to mean all of the software that goes along
> with the kernel to make a usable operating system.

I should think everyone uses Linux to mean the whole OS (kernel +
utils). When I say I run SunOS, BSDI & Irix, I certainly don't mean I
just have the kernels. I have 3 individual setups, each of which works

> Many of the users of Linux have got their idea of the meaning of
> "operating system" from there. That explanation distinguishes clearly
> between the kernel and the whole collection of software you need to do
> actual work, but it suggests calling them both "Linux".
> So it appears that "operating system" is ambiguous. It can mean the
> whole collection of system software, or it can mean just the kernel.
> This ambiguity probably tends to encourage confusion between those two
> different entities.

It really shouldn't be ambiguous. I think everyone reading this knows the
difference in meaning between "kernel" and "operating system"; let's not get
bogged down in semantics, regardless of what some academics say. No matter
what we call it, we should let "the Linux operating system" mean the kernel
that manages the hardware & software tasks, as well as the tools that use the
above through the kernel.

> Most users who use Linux install whole collections of software, which
> contain Linux. These collections include Slackware, Debian,
> Plug-and-Play, RedHat, and so on. They are analogous to
> non-Linux-based collections such as NetBSD, SunOS, HPUX, and Unix. We
> can call such collections operating systems, or ready-to-use
> self-sufficient software distributions, or some other term. Whatever
> name we use, the difference between these collections and the kernel
> is clear.

You make a good point here. The different distributions of Linux are
somewhat like the different variants of Unix out there, in that every one
is just different enough to be different. All the Linux distributions
have a similar code base (Linus' source), as do all the Unix
distributions (whether it be Berkeley or Bell based, it's all related).
I feel we should simply use "operating system" to describe them all.
Microsoft describes DOS, Win95 & WinNT as operating systems; as I
mentioned before, all those utilities are included, just as they are with
variants of Linux/Unix.

> I call these collections "Linux-based GNU systems" to help promote
> unity and cooperation in the whole community. I hope some of you will
> join me in doing this.

This is certainly a noble effort. I, for one, will continue to call it
Linux, if only to save 4 or 5 syllables. However, I will not try to keep
anyone else from calling it what they want. Moreover, I would encourage
Linux developers to work with GNU/FSF developers (and vice versa) as best
they can to promote Linux. The recent discussion of Linux libc vs. GNU libc
has revealed some interesting issues, but we should work to resolve those
issues, not fight about them.

My personal feeling towards GNU/FSF is that they should work at getting
software released a bit faster; say it's a pre-alpha-undeveloped-you'd-be-
crazy-to-use-this version if you want, but let people see it, touch it and
hack at it. I don't mean FSF should work harder or faster; I know they are
some of the hardest working people there are. I just mean that they
shouldn't have to feel like their work is complete before they show it to
anyone. Part of the lure of the latest Linux kernel, that you are as
cutting-edge as can possibly be. Who knows why this is such a lure, but it
is, so we might as well exploit it. Get people using the stuff more
quickly and we can improve it more quickly as well.


Judd Bourgeois
Finger for PGP public key
There's a lost man with a bitter soul
For only a moment did life make him whole
And while he was, he thought he was invincible...
Matthew Sweet, "Smog Moon"