Subject: Re: GPL business practice question
From: "D.V. Henkel-Wallace" <gumby@henkel-wallace.org>
Date: Fri, 18 Jun 1999 13:21:11 -0700

 Fri, 18 Jun 1999 13:21:11 -0700
At 11:40 18-06-99 -0700, Stig Hackvšn wrote:
>I'm looking for examples of companies using the GPL while at the same time NOT
>RELEASING THEIR SOFTWARE PUBLICLY in the way that gpled software is usually
>distributed...
>
>I've heard that there are some very high-value software packages that are like
>this...customers get full source and the GPL, but it's not in their interests
>to redistribute the software, so they don't.  Did this sitation arise because
>of the goodwill and trusting relationship in theis software transactions, or
>was the vendor stuck with the GPL because they used GPLed software in their
>product?

As Kragen says, Cygnus does this but I should emphasize two important 
qualifications.

This is done when the customer is themselves doing something secret (e.g. 
developing a new processor).

The two important qualifications are:
1 - Work that doesn't expose customer-specific details isn't secret.  So 
for example if a processor works well with a particular optimization and 
cygnus develops that optimization that code will go into the egcs releases 
even if the secret processor code doesn't.  Clearly this is good for the 
customer as well because fixes can be provided for the net for that new 
optimization as well.
2 - Cygnus doesn't do this work without a promise that they can release the 
confidential stuff later, either when the processor is released or after 
some timeout, whichever is sooner.  This again is in the customer's 
interest since when they release the processor the details are no longer 
secret and it's in their interest to have the GNU tools available.

Doing things this way has helped make sure that there are GNU tools for 
some processors right from the get-go (like some of the SPARCs for 
instance, or the PlayStation II).

I think we screwed up #2 once, and regretted it, though it turned out to be 
moot since the whole project was cancelled and the changes would have been 
useless (a partial port to a partially-designed experimental processor that 
was never built and didn't look like anything else).  That was almost 10 
years ago, and we learned from it.

The GPL helps here.  It's less common now, but still: sometimes a processor 
developer wants to keep things secret just out of habit and it's good to 
have the GPL to point to to say "well, this is how things have to work."

I should add a disclaimer that since I aren't at Cygnus any more this is 
historical, but I bet that if it had changed I would have heard about it!

-g