Subject: Re: Cygnus and proprietary software
From: kragen@pobox.com (Kragen)
Date: Wed, 31 Dec 1997 06:39:05 -0500 (EST)

On Wed, 24 Dec 1997, Brian Bartholomew wrote:
> > There *is* a Prisoner's Dilemma situation with R&D when you have two
> > companies working on the same software.
> 
> How so?  What's the reward/punishment table?

I looked at it and it's not a prisoner's dilemma, at least in the cases
where you normally would develop software for a business, free or
otherwise.

For n companies working on a particular piece of software, the
payoff of a company x can be approximated as

(sum(i from 1 to n) ef * effort(i)) - effort(x)

where ef is an effectiveness factor representing the ratio between effort
and value produced.

When ef is greater than 1, the more effort you put out, the better off you
are -- and, also, the better off everyone else is.  This is the normal
situation for a business working on software.

When ef is less than 1 but greater than 1/n, the more effort you put out,
the worse off you are, but the better off everyone else is.  Also, the
more effort everyone puts out, the better off everyone is.  This is a true
prisoner's dilemma situation, and while it may happen in the free software
world, it probably doesn't happen in free software businesses.

When ef is less than 1/n, it's not worth it to put any effort into
building the software, so you might as well just forget it altogether. 

(Of course, the above formula is just a rough approximation; value is not
strictly proportional to effort, the value of everyone's effort isn't
equal, the value of my effort to me may be greater than the value of my
effort to you (since I'm solving my problem, not yours), etc.)

I should point out, though, that even in true prisoner's-dilemma
situations, people often have an unfortunate tendency to try to beat each
other instead of cooperating.  Hopefully self-interest will prevent this
in FSBs.

Kragen