[TIP] does licensing of test tools matter?

Fernando Perez fperez.net at gmail.com
Wed Aug 5 15:27:22 PDT 2009

On Wed, Aug 5, 2009 at 12:31 AM, holger krekel<holger at merlinux.eu> wrote:
> Hi TIP,
> i was wondering the other day - is licensing of test libs and
> tools and issue for anyone here?
> As noted elsewhere i am considering using (L)GPL style
> licenses instead of the current MIT license possibly for
> future py.test releases so i am interested in opinions from
> testing people in particular and trust no flame wars arise :)

FWIW, in IPython we use nose and Twisted trial for testing, and just
yesterday I sent an email about py.test to our -dev list (after seeing
your announcement) because I'd like to have a look at it for the
future.  It seems py.test deals with parametric tests much better than
nose (which is already miles ahead of the stdlib), and that's very
important to me, so I got really excited.

But a GPL license would be an issue for IPython, since we are BSD.  In
using a test framework, you end up having to import and call pieces of
it quite often (I have yet to see a test system that's 100% 'external'
for anything non-trivial).  Since, to the best of my understanding,

import foo

carries the fully copyleft effect of the GPL if foo is GPL, then
IPython could not use a GPL py.test without becoming GPL itself.
Since we are not going to relicense IPython as GPL, we could not use

So that's just a data point.  Note that I am not advocating in
principle for or against the GPL, I think it's a perfectly fine
license for many uses.  I am only trying to give you a useful answer
from a reasonably well accepted python project (IPython), which
coincidentally was very interested in py.test just now.

Ultimately you wrote the code, you can choose how to license it, and
I'd rather see you GPL py.test if that is the best choice *for you*.

All the best,


More information about the testing-in-python mailing list