Subject: Re: Would people pay to choose what gets developed next?
From: Ian Lance Taylor <>
Date: 26 Feb 2001 17:28:23 -0800

Brian Behlendorf <> writes:

> > > > BTW, there was an issue with this sort of "directed development" a few
> > > > years ago.  As I recall, RMS was concerned that Cygnus would concentrate
> > > > on things that its customers wanted, to the possible detriment of the
> > > > needs of the rest of the user community.  I think that Cygnus sailed
> > > > through that moral quagmire pretty cleanly, but others might not.
> > >
> > > I don't see any moral quagmire - developers & their communities are free
> > > to scratch their own itches only, and are under no obligation to scratch
> > > someone else's.
> >
> > There was a potential moral quagmire, because gcc is a fairly complex
> > program.  It was entirely possible for Cygnus to add complexities to
> > the compiler which were geared to specific customer requirements, but
> > which made other ports more difficult.
> But that ignores Cygnus's interest which is in getting more gnu tools port
> business, thus making it less portable would work against them.

Ah, but making the compiler harder to port by increasing the internal
complexity would actually be an advantage for Cygnus, as it would
raise the barrier to entry for direct competitors.

This isn't purely theoretical.  I honestly think that the significant
complexity of BFD and the GNU binutils, which derived mainly from the
coding style and rather than anything inherent in the problem domain,
actually did make it harder for people to compete against Cygnus.  BFD
was written at Cygnus, and for several years every single person who
really understood BFD worked for Cygnus.  (There are now a few other
people, not at Cygnus and not ex-Cygnites, who have climbed the steep
learning curve.)

> The other alternative is the typical one: if you don't like it,
> submit patches, or if they're too quickly dismissed, fork.

Oh, sure it is not an absolute problem in any sense.  It is only a
matter of raising costs.

> > In general I agree that Cygnus handled the quagmire well, because a
> > large percentage of the engineering staff was committed to the overall
> > goals of free software in general and the GNU project in specific.
> Even if they're only committed to the long-term health of the GNU toolset,
> aside from the GNU project or the FSF, their interests would be aligned
> with the community's.