BAD recap

Robin Rowe rower@MovieEditor.com
Thu, 10 Jan 2002 10:07:13 -0800


Hi. There were about twelve of us at Au Coquelet last night. Besides those
who you might expect, the group included an anthropologist, a Kernal Cousin
journalist, and an EFF representative -- an interesting mix! Below is a
recap of some of the discussion. I've omitted names to protect the innocent.

Not everyone agreed when I suggested that it would be good to encourage and
support Windows ports of open source software (the topic of my current
article in Linux Journal). Some thought Windows ports should be deliberately
unsupported by OSS projects as a snub to Microsoft, or that working with
Windows programmers or users was too unpalatable to consider anyway.

A separate discussion followed, debating the significance of the difference
between "free software" and "open source." I said I usually use the term
open source, because as a programmer my primary concern is access to the
source. I also said that when someone says free software I'm not sure what
he or she means. I would have to ask does the person mean GPL, BSD, or
public domain with or without source code? But, another opinion was that
free software always means Richard Stallman's definition, that there is no
confusion. To me that sounded a bit like saying windows always means
Microsoft Windows.

There was also an interesting difference of opinion on what constitutes
software freedom. I said that I find the GPL more restrictive than the BSD
license. My reasoning is that under the GPL I waive some rights to my own
works. In other words, I have to give my source code back. The counter
argument was that under the BSD license I waive any rights to others'
derivative works based on my work. I pointed out that in my opinion as a
commercial programmer the GPL seems harder to get a commercial sponsor to
embrace. That's because a company often wants to defer the final marketing
decisions until after the code is completed, to have the greatest
flexibility to react to future business conditions. However, my reasoning
was deemed too practical, apparently not in the spirit of free software.

An autoconf question came up. A programmer trying to build an existing OSS
project described his trouble getting the KDE includes to be found during
compile because the makefiles don't point to the right location. I suggested
that once he'd hacked the makefiles to fix the immediate problem he should
revise the input files to configure and run aclocal and what not to generate
the correct configure file, not just leave that broken. That way configure
will generate the correct makefiles. I've done that with many OSS projects
I've built because the original developer missed some configuration checks
that didn't effect his system. It is easy for that to happen because many
OSS projects don't get built by many people besides the original author.
Some don't realize that in an OSS project the setups for configure are
there, that you should fix those because configure is a generated file.
Generating a new configure is the right way to go, especially if a project
has makefiles in many directories making manual changes time consuming.
Autoconf documentation is available on the GNU Web site. No need to buy a
book.

Problems with autoconf, make, or C++ are easier to deal with if others help.
Because I like working on OSS I'm thinking of forming a group locally for
programmers working on open source applications or who would like to but
haven't been able to get started. If anyone is interested please send me a
note off-list.

Cheers,

Robin