Subject: Re: Studies
From: "Adam J. Richter" <>
Date: Sun, 30 Nov 1997 19:10:29 -0800

	It's nice to see some new people and new subjects on the fsb
mailing list.  I agree with much of what Joel is saying.  Here are
some additional comments on it.

Joel Weber writes:
>> = ?
>  = Joel Weber (  
>>Someone said that one of the big differences between the GNU developer
>>community and the Linux developer community is their attitude toward
>>commercial software.  If I understand correctly, the FSF was founded on
>>the premise that proprietary software is immoral, and should be stamped
>>out, if possible.  This attitude seems to be almost nonexistent among
>>people whose introduction to free software was Linux.


	Yes, and it is unfortunate that a large number of users of the
Linux kernel do not understand what FSF stands for and do not
appreciate that general GNU software is an order of magnitude more
code than the Linux kernel and Linux-specific software, and most of
the experience that they call "Linux" is really GNU + X windows +
other free software not written specifically for the Linux kernel, in
that order.  On the other hand, much GNU software was written by
people who do not fully support the "proprietary software is immoral"
view but rather write free software for all of the other reasons that
people write free software.  (When I say "free software", I mean "free
software" in the GNU sense unless otherwise noted.)

	Really, the Linux developer community is not separate from the
GNU developer community.  It is the same bunch of people contributing
to the same pool of software.  It is just unfortunate that much of the
Linux and BSD *user* communities underestimate the contribution of the
GNU project, although I think this problem is being reduced these
days, and will continue to be reduced with widespread adoption of
version 2 of the GNU C Library by all major (GNU/)Linux distributions
in the next few months, resulting in less identification with "Linux"
and more with "GNU."

>But I often wonder why it is that people seem to think that the Linux kernel
>is the most woderful thing in the world.  It is certainly a useful program,
>but I doubt it is any significantly better than the *BSD kernels.  Having
>a goal of simply running everything on one particular kernel is not really
>worth gettingg into religious wars about.

	I agree that there is no need to convert Hurd or
{Free,Open,Net}BSD users to Linux kernel users.  We're all
contributing to and using the same software pool, and I think the
benefits of some collegial internal competition outweigh the
relatively small and shrinking costs of porting between these very
similar systems.

	On the other hand, I do think that is useful to be able to run
as many different types of software as possible without rebooting or
shutting down other programs that are running and to support as much
hardware as possible.  (I am not sure if you actually meant to say
that that should be a lower priority.)  For example, I think that
solid Windows emulation is the next big step for Linux and GNU systems
in general.

>I think that having the source code for all the software on your computer
>and being allowed to freely share and change it is a much more worthwhile

>If anything, the freebsd community may care more about having the source
>available than the Linux community does.  Apparently, you can rebuild the
>entire freebsd system by doing `make world'.  I'm sure you can't do that with
>Slackware, and I doubt that it is easy with the other dbistrstributions,
>either.  I know that RMS wanted debian to come with biaries that were
>compiled with the -g flag so they could be debugged, and the feeling of the
>debian developers was that it would make the system bloated and most users
>wouldn't know how to use a debugger...

	Every release of Yggdrasil Plug & Play Linux since 1993 has had
a fully buildable source tree.  If you installed the entire system and
the source tree, you could rebuild it by typing and crossing enough
of your fingers:

		cd /usr/src
		make all
		make install

	With the free portions of Red Hat and the latest experimental
Debian, there are source packages in a well defined format that would
allow you to write a small shell script to regenerate each package
from scratch.

>Another difference is that the GNU community tends to try to write its
>userland so that you can run it on any Unix kernel.  This often turns
>out to be useful if, for example, you have an hp300 running netbsd, and
>you like the improved syntax that GNU find supports.

	autoconf is a huge under appreciated contribution to free
software by The GNU Project.  Autoconf makes it easy enough to make
code portable so that many more free software authors do so, greatly
increasing the reach of free software.  Also, learning to use the
relatively uniform "./configure" scripts is a relatively small time
investment with the payoff of enabling users to build a lot more
software from source, and more easily.  Lots of developers are
adopting, not just the GNU project.  Fortunately, there is no prejudice
that I am aware of against autoconf specifically among Linux users
or developers.

Adam J. Richter     __     ______________   4880 Stevens Creek Blvd, Suite 205     \ /                  San Jose, California 95129-1034
+1 408 261-6630         | g g d r a s i l   United States of America
fax +1 408 261-6631      "Free Software For The Rest Of Us."