[License-discuss] Ethical open source licensing - Persona non Grata Preamble

Brian Behlendorf brian at behlendorf.com
Fri Feb 21 19:26:25 UTC 2020

On Fri, 21 Feb 2020, Jim Jagielski wrote:
>> On Feb 20, 2020, at 9:16 PM, Eric Schultz <eric at wwahammy.com> wrote:
>> Instead I've been thinking through ways in which licensing and ethical FOSS community policies can interact in order to discourage and shame morally corrupt users.
> So who defines who is "morally corrupt"? Who exactly is a "bad actor"? What is "evil"?

The licensors define that, in their preamble. You don't have to agree.

I would likely never use such a license on my software, I would strongly 
avoid writing code that depended upon such a license when writing my own, 
and I would probably prefer to use software licensed otherwise.

But, that is true for many other certified Open Source licenses. 
Likelihood of widespread adoption has never been a formal criteria for 
certification, aside from proliferation and inadvertant duplication 
concerns. Back when the open source software world was much smaller, and 
we were still struggling for acceptance, and we really did want every 
company large and small to be a part of the revolution, then a big tent 
was called for by some (permissive) and a principled stand by others 
(copyleft). This is a different sort of principled stand, one likely to 
repel even more people than copyleft depending on the non Gratas, but I 
struggle to see how it's any less valid than copyleft.

I am struck by the brilliance of the self-shaming aspect of this - that 
ICE would be free to use this software, but would have to redistribute 
ICE-shaming sentence if they redistributed or derived works from it. 
Unless I'm misunderstanding though, such text would not have to be 
displayed in the user interface or if deployed over the network, only 
where such copyright notices are usually displayed. Rarely does that make 
its way to end-users. So there's a flaw in this that may require 
patching-over with an AGPL-like requirement to display or offer the 
license text in any end-user UI. But even then, that's not disqualifying.

Clearly this is not a license you would expect to see mass adopted for 
lower levels of the stack - operating systems, network daemons, databases 
- or mainstram end-user software like a web browser. But if I had just 
pulled together (through a training set + algorithms) a pretty great 
facial recognition system, I'd give this license serious thought - 
especially if I thought I could get a critical mass of other 
facial-recognition developers & researchers on board with a particular set 
of non-Gratas, such that an alternate engine more permissively licensed 
would struggle to compete. But I'd have to also concede that Bad Actors 
are unlikely to be persuaded by (or even read?) the terms of such a 
license, or would figure out how to route around any redisplay 
requirements. But the symbolism and signalling would be compelling, a step 
up from a bumper sticker or t-shirt.

One more practical and probably negative impact: it will always be easier 
to add new names than to remove old ones, because adding new names is a 
simple sublicense that can be done by anyone, but removing a name will 
require a relicense with the consent of every contributor who contributed 
under that license. So, in my facial-recognition example, I would likely 
as for assignment of the right to relicense from any contributor, so that 
I might be able to remove a name once they've cleaned up their act. 
Otherwise a list that can only grow becomes an embarrassment and 
ineffective at actually changing behavior, it just becomes a howl in the 
night. But centralizing IP (rather than a mere right to redistribute a la 
DCO / CLA) is something I think we've done well to avoid.


More information about the License-discuss mailing list