speeding up Octave development (was: Re: m-code)

[John W. Eaton]

On  4-Mar-1999, address@hidden <address@hidden> wrote:

| I agree that there is a need to lighten jwe's load
| w.r.t. to development.  Having a cvs tree would be 
| very helpful.

An anonymous CVS archive could be the catalyst that ignites an
explosion in the number of contributions, but I'm not holding my
breath.  Don't get me wrong -- I'm willing to try it out, and I agree
that anonymous access to the latest sources via CVS might help some
people, but I think there are several other things that can speed
development even more.

One thing that would help a lot would be a relatively small amount
of funding to make more of my time available for working on Octave
instead of other projects.  When I'm busy doing other things (which is
necessary from time to time in order for me to keep my job) progress
on Octave seems to stop.  Without some funding to free up more of my
time for Octave, I don't think this will change any time soon.  The
reason is that no one else has been doing enough work with the core of
Octave to really be able to take over maintaining it.

Another thing that would increase the speed of development is
*quality* contributions of *actual code* from many different people.
This includes bug reports with patches, not just an expectation that I
will fix the problem once it is reported.  Also, it is not enough to
submit long wish lists or partially and poorly implemented code.  It
takes too much time to clean things up to include them.

Eric Raymond has received a lot of attention for saying that the
`bazaar' method of free software development can make a project move
much faster.  He makes some compelling arguments, but he based his
conclusions on a relatively small sample of projects, and most of them
are projects that competent hackers think are cool.  Now I see people
claiming that progress on any project will be faster if it switches to
bazaar mode.  I just don't think it's true.

I think Octave's development is rather open, but progress has also
been relatively slow at times.  I certainly don't think we are in
`cathedral' mode.  There aren't any big secrets here.  Test releases
and snapshots of the development sources are publicly available (they
are not made automatically at random points in time because I don't
think that is really useful for most people, even other developers).
If you want to be on the octave-maintainers list, you can subscribe.
There are currently only 16 people subscribed and the traffic on that
list is nearly zero.

My experience has been that, for whatever reason, hackers seem to be
excited by C compilers, kernels, networking code, gui toolkits,
editors, and even some general purpose languages like Perl.  But they
don't seem to be too excited by spreadsheets or programs for matrix
algebra.  At the same time, I've seen lots of physicists and
mathematicians take an interest in Octave, but most of them are
interested in using it to solve some problem, not in extending it.  If
they do write new functions, it seems that they view their code as
something that not many people would be interested in, or that is not
polished enough for distribution (they are probably often right about
that :-), so they don't contribute their code.

I could go on, but I'll just point out an essay on this topic that
expresses ideas that are fairly close to what I've been thinking for a
while now:

  http://www.its.caltech.edu/~jafl/misc/bazaar_fizzle_fail.html

It was mentioned on this list a few weeks ago by Stefano Ghirlanda
<address@hidden> (thanks!).  It's definitely worth reading if you
are at all interested in what can make some software projects take off
(Linux, Perl, the Gimp) and why others don't seem to go nearly as
fast.

jwe