[License-review] For Approval: OpenMDW License Agreement, versions 1.1 (OpenMDW-1.1)

[Shuji Sado]

Carlo-san and All,

The discussion seems to have branched into several issues again, but I
think Carlo-san's message connects them quite well.

First, I agree with Carlo-san that a plain software test is not enough.
That is exactly why I raised the analogy with the OFL and the IPA Font
License in my previous message.

If OpenMDW is approved by OSI, I think it would naturally belong in the
Special Purpose category. I do not mean that an AI-related license should
be reviewed under a different or looser OSD standard. Rather, while
applying the same OSD, we also need to recognize what constitutes a natural
licensed unit in that particular domain. The history of the font license
reviews shows that OSI has already had to deal with this kind of question.

I also agree that it would be useful to make explicit that rights properly
granted to downstream recipients survive the termination of an upstream
licensee's rights. That seems to me a useful clarification and improvement
to the license.

On the other hand, regarding Carlo-san's phrase "general waiver of one's
right to the licensor," if "waiver" is meant literally, I am not sure that
it accurately describes the structure of OpenMDW.
OpenMDW does not require the licensee to waive the copyright claim itself,
nor does it prohibit filing a lawsuit. The licensee remains free to bring
the claim. The consequence is that the grants received under OpenMDW
terminate. Of course, whether that termination has the practical effect of
discouraging the exercise of rights is a separate question, but I think
waiver and termination should be distinguished.

There is also the OSI-approved OCLC Research Public License 2.0, which uses
an action for "intellectual property infringement" with respect to the
Program as a termination trigger. That language is at least broad enough to
include a copyright infringement claim.

Seen in that light, I think the question may be moving away from whether
copyright retaliation is categorically impermissible in an Open Source
license, and closer to the question of proportionality: to what licensed
unit should the termination apply? In other words, this seems to connect
back to the earlier question of whether the Model Materials are
appropriately defined and bounded as a unit.

Finally, on Software Freedom.

This discussion prompted me to read the License Review Process again, and I
confirmed that software freedom is expressly mentioned alongside the OSD. I
therefore need to revise part of my earlier understanding on that point.

However, the Review Process does not define the OSD as "merely a means."
The OSD provides a concrete and reasonably predictable operational standard
for both license drafters and the review community. I think software
freedom is better understood as a backstop for cases where a license
formally satisfies the OSD while effectively circumventing its purpose.

If software freedom were instead to become a free-standing balancing test
under which a license could be rejected based on an open-ended policy
judgment, license drafters would no longer be able to know in advance what
is acceptable, and much of the predictability provided by the OSD would be
lost.

So I am not rejecting software freedom as a consideration. I am only
cautious about treating it as an unlimited independent policy test
alongside the OSD.


--
Shuji Sado
Chairman, Open Source Group Japan
https://opensource.jp/
English blog: https://shujisado.org/
Japanese blog: https://shujisado.com/


2026/8/31 22:16 Carlo Piana via License-review <
license-review at lists.opensource.org>:

>
>
> ------------------------------
>
> *Da: *"Michael Dolan" <mdolan at linuxfoundation.org>
> *A: *"License submissions for OSI review" <
> license-review at lists.opensource.org>
> *Inviato: *Venerdì, 28 agosto 2026 23:15:05
> *Oggetto: *Re: [License-review] For Approval: OpenMDW License Agreement,
> versions 1.1 (OpenMDW-1.1)
>
> On Fri, Aug 28, 2026 at 5:06 PM Richard Fontana via License-review <
> license-review at lists.opensource.org> wrote:
>
>>
>> Yes. Incidentally, the OSI should bear in mind that OpenMDW-1.1
>> functions as a "plain old software license" because it is designed in
>> part to be used for software. I might view it somewhat differently if
>> it was presented as, say, a "weights license" only (perhaps submitted
>> as a special sort of "AI license" under what I gather was contemplated
>> when the OSAID was adopted). I don't think the OSI should be
>> classifying this as an "AI license" and adopting some new, presumably
>> looser, standard of approval for it based on that classification. It
>> is by design not a license just for models in a narrow sense, it is a
>> license for code as well as other stuff. If it is approved, someone
>> could conceivably use it in a completely non-AI-related context
>> involving just code.
>>
>
> Richard, this is a good point, and I'd like to expand on it a bit.
> OpenMDW-1.1 was submitted for review as a software license, under the same
> standard as any other license OSI has reviewed. We are not asking for an
> AI-specific classification or a looser OSD interpretation. Your observation
> about non-AI reuse is correct and by design: inference code, preprocessing
> tools, or other software licensed under OpenMDW-1.1 as part of one release
> can be extracted and reused by developers with entirely different model
> weights that are under other licenses, or with no model at all. Each
> component published under the license carries the OpenMDW grant, and the
> license needs to work as a plain software license in that very scenario.
>
> Mike
>
>
> Mike, Richard, all,
>
> I hesitate chiming in here because I don't want to compromise my judgment
> by taking a too early position on the very clause under scrutiny, but this
> exchange is just too inviting to remain silent any longer. Caveat: this
> entire message is a tentative position and does not necessarily reflect
> what the Licensing Committee or the Board would decide.
>
> I partially disagree with one premise. We cannot ignore that the license,
> even if applied to software, would have a direct legal/licensing impact on
> related material *outside* what we consider "software". Software and non
> software subject matters would have an impact on each other, even if each
> one can be extracted and reused under the same license in a different
> context. Therefore we cannot just pretend such a license is a pure software
> one and approve it on that basis alone. Nor can we apply a looser standard
> just because it's AI (which includes software, so this would be
> contradictory) but there seems to be no disagreement here.
>
> As a general statement (I), I am really wary of porting the same reasoning
> we apply to software to an entirely new and broader category of artifacts.
> This would always  require a "ceteris paribus" (= all other things being
> equal) premise. I would probably not have issues with this clause if
> applied to software in both source and object code. I am struggling to find
> a solution to the conundrum at hand, but one thing is clear to me: we
> cannot just simply apply the "would this be acceptable for software" test.
>
> As a general statement (II), I would submit at least two requirements:
>
> a) all downstream recipient of the Model Materials should not be affected
> by the termination of the license if they are not directly or indirectly
> part of the same litigation. This may be already clear to many, but the
> license does not address this and under many laws the termination of the
> license would risk to terminate the downstream relicensing, unless a clause
> similar to the GPLv2's §6 first period existed. I would personally suggest
> to add something to that effect so any doubt is removed.
> b) while a litigation retaliation clause has in the past been considered
> by and large both compliant and ethical, IMVHO the retaliation must be
> proportionate; therefore it should not transform into a more or less
> general waiver of one's right to the licensor as a price to become a
> licensee. This is distilled from a ius gentium principle, if and how it
> applies to OSD can be discussed separately.
>
> In the thread we have been submitted a lot of paradoxical situations
> discussing b) some of which probably too far fetched, but not all. I wonder
> therefore if there would be any way to restrict the effect of the
> termination to the subject matter of the litigation? Eg.: software is
> claimed to infringe, the license on the software and on the model that same
> software supports is terminated, but not the license of the documentation,
> of the training material and information, of other unrelated software
> artifacts and models etc.?
>
> As a general statement (III), finally, the licensing is only a
> pre-condition for something being Open Source. More is required. Likewise,
> we must not include too many other hypotheticals: for sure me must exclude
> the intent of the drafter. Only the operation of the license if *per se* it
> violates OSD *and/or* impairs the Software Freedom principles. OSD is a
> means to an end, the end is Software Freedom, that at least is my take. I
> have a feeling in the AI realm licensing will have a much, much weaker
> impact than in the traditional software one, anyway.
>
> Again, it is very hard to decide. Whichever decision we make we would
> probably err on one side, not that this would be a reason not to decide.
>
> Thank you for a thought provoking submission causing a very lively
> discussion upon licensing -- something we haven't seen in years, I reckon.
>
> Cheers
>
> Carlo
>
>
>
> _______________________________________________
> The opinions expressed in this email are those of the sender and not
> necessarily those of the Open Source Initiative. Communication from the
> Open Source Initiative will be sent from an opensource.org email address.
>
> License-review mailing list
> License-review at lists.opensource.org
>
> http://lists.opensource.org/mailman/listinfo/license-review_lists.opensource.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensource.org/pipermail/license-review_lists.opensource.org/attachments/20260901/d981af37/attachment-0001.htm>