Subject: Re: Exploring the limits of free software
From: "D. V. Henkel-Wallace" <>
Date: Wed, 26 May 1999 08:06:37 -0700

At 16:25 26-05-99 +0900, Stephen J. Turnbull wrote:
> >>>>> "g" == D V Henkel-Wallace <> writes:
>None of these options should involve much additional expense, unless
>you bought crummy work in the first place.

Ahh yes, but we live in an imperfect market, neh?

Most people have no way of evaluating the quality of the code they receive, 
except "does it do the narrow thing I asked for."  They certainly have 
little way to do so beforehand.

This does give an advantage to the incumbent.  It's branding.

At 00:25 26-05-99 -0700, Brian Behlendorf wrote:
>Based on my experience in the Apache project, the onus is on the submitter
>to create patches acceptable to the rest of the group.  That could mean
>follow-on work needs to be performed and paid for; in that case, you look
>for sponsors who would benefit from Linux thread support in Cygnus tools.

Unfortunately this is simplistic.  The majority of users are Linux users; 
they get these patches with their linux distributions and don't understand 
(for the reasons above) why they can't "just upgrade" their gdb when a new 
one comes out.  This same situation is why there are linux-specific 
releases of almost all the tools.

Fortunately we can actually do an experiment.  Since we have multiple bid 
sites, one can handle _only_ projects which are approved by a board and for 
which some percentage of the final payment depends on integration (can't be 
a small percentage or you'll just get inflation; people will work for the 
80% and just factor in losing the final 20%).  Another board can take any 
old project.  Let's see what happens.

Wil gresham's law of programming drive out the good projects?