[License-review] Request for Open Constitution License v1 for Approval

Josh Berkus josh at berkus.org
Fri Aug 18 01:04:39 UTC 2023


Thanks for submitting this!  We appreciate the effort to create a 
license specifically in the AI realm.

I'll note that overall, this license seems to have been created by 
incrementally editing your prior terms of service, without a review to 
ensure that the resulting document was internally consistent and clear. 
The suggestions below are in pursuit of improving this license to the 
point where it can be acceptable.

There are also numerous (i) callouts in the license.  It is not clear 
whether these are part of the license text, or are intended to be 
explanitory.  The text in many of these callouts directly contradicts 
text elsewhere.

On 8/17/23 10:38, Kevin P. Fleming wrote:
> 
> This license contains a definition of "Open Source" which is not based 
> on the OSD. It seems highly unlikely that the OSI would approve a 
> license containing a conflicting definition of this fundamental term.
> 
> Article 1 section b is somewhat poorly drafted, but in any case requires 
> that a 'derivative work' comply with a third party's "Acceptable Usage 
> Conditions". This is clearly not compatible with OSD 5 and 6, which 
> disallow discrimination against person(s) or group(s), and against 
> fields of endeavor. Article 5 is also incompatible with OSD 5 and 6.

Yes, this.  Licenses cannot include, by implication, external terms and 
conditions. Such conditions must be part of the license, and they must 
comply with the OSD.

> 
> Article 2.1 is not compatible with OSD 10; the license cannot dictate 
> any specific technological mechanisms.

Also, 2.1 conflicts directly with the subsection "Open Constitution 
License Is technology-neutral."  And with the very next paragraph under 2.1.

Also, the whole business of machine vs. human-readable source objects 
very hard to understand, and I suspect the result of further edit conflicts.

Article 3: Tokenization: this seems to say that licensees must have some 
kind of specific electronic ID token to be licensees?  Is this an 
AI-specific thing?  Because as a general software dev, it's completely 
unclear to me what this section is getting at.  This seems to be 
intended to be in support of Article 4, which would be related to the 
open sourcing of AI data models, but the language is not at all clear.

Article 5: This makes it clear that the idea is for Acceptable Usage 
Conditions to be something that any licensor can do for their individual 
source object.  This isn't going to work, not if you want this to be 
open source.  It's also unenforceable and impossible to comply with.

It also adds a dependancy on yet another external document, the Global 
Statutes of the Open Constitution Network, which isn't going to work for 
the reasons explained above.

So, interesting first draft, hopefully we can edit this into a place 
where it's actually open source.

-- 
Josh Berkus





More information about the License-review mailing list