[alexb at ufl.edu: Re: support requirement]

[Chris F Clark]

> I don't see any gotcha.
>
> I would probably not need Vendor X's documentation and/or output
> to reimplement the program.

You might if they were sufficiently clever.  I can think of several
pieces of software where the vendor has managed to implement something
in a non-obvious way that has significantly improved performance over
the naive implementation and that there are no competitive free
software clones of for that reason (at least none that I know of and
I've looked in some cases).  If you can't think of any non-free
software for which there are no free alternatives, you are missing a
lot of software (and perhaps you are lucky enough not to need it).

The best example I can think of comes from the EDA industry.  There
are a handful of commmercial simulator vendors that dominate the
market and charge big-bucks for their tools, and have done so for some
time.  As far as I can tell, there is no effective free software in
that market--and that's despite the fact that there are standard
documents describing the simulation languages, which should mean that
someone could just "implement the spec" and come up with a free tool.

I think part of the reason is that the obvious implementation performs
too poorly (hours of simulation time versus minutes) and the vendors
have carefully tied their customers hands with NDA's so that the
customers can't go to a free software developer and say "we would like
you to make a free version with the following optimization" since the
knowledge of the optimization is covered by the NDA.

-Chris