Re: Ada policy

[Robert Dewar <dewar at gnat dot com>]

Per Bothner wrote:

> Richard Kenner wrote:
>
> > As I'm sure you know, if you start with copyrighted code and "trim it
> > down", you still have copyrighted code.  Making a non-proprietary
> > version of a test case involves a lot more work than that.

> A trimmed-out test case, with comments removed, and identifiers renamed,
> is still a "derived" work, but presumably many customers would be
> willing to disclaim copyright interest.

This might work in some cases, but it's much harder than you think.
Every such case has to go to company lawyers, whose job it is to
protect the IPR interests of the company, and who naturally default
to a conservative "protect everything" viewpoint. In the absence
of a very strong argument of value to the company, it is hard to
get such agreements. Furthermore, we have been very successful
in persuading companies large and small to contribute substantial
chunks of proprietary code, under the absolute guarantee of
fierce protection. That's lead to us having a very effective
suite of real code with tens of thousands of files, and many
millions of lines of real Ada application code. So we have to
be careful not to confuse the message.

> Perhaps as part of your
> bug resolution procedure you could ask "could you please review
> this simplified test case as to whether we can release it".  This
> does involve some customer education, so it is probably only worthwhile
> for your bigger customers - i.e. those who submit lots of test cases.

Right, it might work in some limited number of cases. We have some
experience with this when it comes to a few cases in which customers
have contributed code. Even that can be a tricky case to deal with
and very often we have ended up just rewriting from scratch rather
than deal with the legal problems. We don't get very much contributed
code of this type, so these are isolated cases from a small number of
customers, and yes, education is a large part of the job here.