Date
1 - 12 of 12
Commutativity of SPDX expressions
|
Richard Fontana
I'm working on some draft documentation for Fedora around use of SPDX
expressions in RPM spec file License: fields. I was surprised to apparently not see anything in the SPDX spec that says that the AND and OR operators are commutative. I want to assert that the expression "MIT AND Apache-2.0" is equivalent to "Apache-2.0 AND MIT". Does the SPDX spec actually take no position on this? Richard |
|
|
|
McCoy Smith
At the risk of sounding like I’m hijacking this to re-raise my prior issue:
toggle quoted message
Show quoted text
If AND is the operator to be used when having different inbound vs outbound, then AND may not be commutative, since the order of listing the licenses may convey information about which license is inbound vs outbound, and (maybe) which license applies to different parts of the code. Which militates to me toward a new expression, but I’ve made that point already. On Jul 17, 2022, at 11:22 AM, Richard Fontana <rfontana@...> wrote: |
|
|
|
Warner Losh
Please define "INBOUND" and "OUTBOUND" licenses. None of the open source licenses indexed by SPDX grant permission to relicense the derived work, so any work including them either is solely the original license, or an AND of the project's license and the original license (to the extent the project creates a derived work). This is a clear case for "AND" where both licenses must be complied with. SPDX expressions do not have additional , secondary meanings, like what is the preferred license, what license do we accept changes for this derived work under, etc. All those secondary meanings would, if any, would need to be spelled out explicitly by the project wishing for them to apply for that project. For example, if a project takes MIT licensed code and creates a derived work they wish to license as GPL, they must list the license as GPL AND MIT because their desire to license under the GPL does not override the MIT license nor the IP of the original author. To say that the GPL is a proper-superset of the MIT license would be taking a legal position that's more aggressive than has been taken in the past, and I don't think that the SPDX project wants to be in the position of creating a matrix that says what is a superset of what so you can 'reduce' the complexity of the SPDX-License-Expression using some set of rules. As for "AND": IThe standard says the following: 'If required to simultaneously comply with two or more licenses, use the conjunctive binary "AND"..." Since each license is given equal weight in this construct, I'd say that AND is commutative where A AND B == B AND A. The standard has no verbage to the contrary to suggest one is controlling when there's a conflict, for example. Nor any of the other problems that might result from them being non-commutative. The standard is not explicit that it is commutative, but common usage for AND is. In the absence of a more specific definition, common usage would dictate interpretation should this be litigated. I'll let the better legal minds here, though, chime in as to whether or not the SPDX standard, as written, would need any correction to its language to address this potential ambiguity. Warner On Sun, Jul 17, 2022 at 1:18 PM McCoy Smith <mccoy@...> wrote: At the risk of sounding like I’m hijacking this to re-raise my prior issue: |
|
|
|
J Lovejoy
Hi Richard,
toggle quoted message
Show quoted text
Annex D explains the order of precedence for the operators and use of parentheses. https://spdx.github.io/spdx-spec/SPDX-license-expressions/ I admit, I find the use of parentheses easier to understand overall (than relying on remembering the order of precedence). I’m not sure it explicitly states that "MIT AND Apache-2.0" is equivalent to "Apache-2.0 AND MIT” but I think that’s kind of implicit, no? I also think this entire annex could use a re-write to make it a bit more user-friendly (on the topic of improving documentation…) Jilayne On Jul 17, 2022, at 12:21 PM, Richard Fontana <rfontana@...> wrote: |
|
|
|
J Lovejoy
Hi McCoy,
toggle quoted message
Show quoted text
I’m wondering if you are trying to adapt SPDX identifiers in a situation not anticipated. Consider that aim of an SPDX document (as per the SPDX specification, and thus, using SPDX license ids in the various specification field, is to communicate licensing, copyright, provenance, etc. information for a given bundle of software. For example, I sell you Jilaynes-awesome-software-app and provide an SPDX document for that software product. The licensing info in this context would be presubaly what I think you are referring to as the “outbound” license - that is the license under which the software is used by the recipient. Let’s say, Jilaynes-awesome-software-app includes some open source software under various open source licenses, say, MIT and Apache-2.0, and I also added some of my own (new) code under BSD-3-Clause, that all of this can be reflected in the appropriate license fields at the package, file, and/or snippet level. I think of “inbound”, in relation to open source software, as usually referring to the license under which contributions are provided to the project. But I think you might be meaning “inbound” in relation to Jilayne’s-awesome-software-app - that is, the open source software that I incorporate into my app under MIT and Apache-2.0. Is that right? Thanks, Jilayne On Jul 17, 2022, at 1:18 PM, McCoy Smith <mccoy@...> wrote: |
|
|
|
McCoy Smith
Rather than getting into further debates about what various licenses do and don't require, or for that matter what copyright law does or doesn't require, I guess I'd turn back to the ath5k example.
toggle quoted message
Show quoted text
Is the license designation they used the same as the AND operator in SPDX? I think it is not (or if AND encompasses it, AND may be interpreted too broadly so as to potentially cause confusion or incorrect assumptions about the license state). Ath5k license designation is here: https://lwn.net/Articles/247806/ Now, people are free to respond back that the ath5k license designation is legally invalid, but I for one will not stand here and have Richard Fontana's legal skills besmirched! -----Original Message----- |
|
|
|
Richard Fontana
The order of operations is a different issue, I think. I guess the
toggle quoted message
Show quoted text
SPDX spec assumes, as you say, that commutativity of AND and OR is implicit (like counterpart operations in propositional logic), but this implicit property was not obvious to one Fedora contributor. Richard On Sun, Jul 17, 2022 at 4:08 PM J Lovejoy <opensource@...> wrote:
|
|
|
|
Warner Losh
On Sun, Jul 17, 2022 at 2:43 PM McCoy Smith <mccoy@...> wrote: Rather than getting into further debates about what various licenses do and don't require, or for that matter what copyright law does or doesn't require, I guess I'd turn back to the ath5k example. Each of the individual files retains the original copyright and license, as the original author required. You are required to still abide by the terms in those files (but each individual grant is not the sum of the requirements). The current kernel.org Linux ath5k driver is marked as 'MODULE_LICENSE("Dual BSD/GPL");', The kernel.org version of this driver does not have these changes included. In addition, the OpenBSD folks were none-too-happy with this attempt to strip off the BSD licenses. https://undeadly.org/cgi?action=article&sid=20070829001634 has the details (but Google finds many other instances, I've not chased them all down). LICENSE_MODULE is beyond the scope of SPDX and is up to the Linux Kernel Community what licenses they support and when. The SPDX matching tool, which implements the SDPX license matching guidelines, would say that there's multiple licenses you must comply with. That means the union of all the licenses which is the meaning of AND in a SPDX-License-Identifier which I believe would be the result for several of the files. I've not run it on the current version of these files, but have obtained that result for other code that has multiple licenses. I'm not entirely sure, given the contentious history that this makes a good example, though. Warner > -----Original Message----- |
|
|
|
Gary O'Neall
I've always assumed the AND and OR operators to be commutative and the SPDX Java tools take full advantage of the commutative properties when comparing license expressions.
toggle quoted message
Show quoted text
I would welcome a pull request to Annex D to clarify this since at least one member of the community found this ambiguous and/or confusing. Gary -----Original Message----- |
|
|
|
Dear Gary,
I've always assumed the AND and OR operators to be commutative andI've made a pull request for this :) Not sure whether it'll make 2.3 though - has the final draft been released yet? https://github.com/spdx/spdx-spec/pull/748 Best wishes, Sebastian |
|
|
|
Gary O'Neall
Thanks Sebastian - since we haven't finished the review of the version 2.3,
toggle quoted message
Show quoted text
I think there is still time. Best regards, Gary -----Original Message-----though - has the final draft been released yet? |
|
|
|
Richard Fontana
I feel like what some projects might find useful is something like:
toggle quoted message
Show quoted text
SPDX-License-Identifier-Concluding-What's-Been-Contributed-As-Of-Some-Past-Time: SPDX-License-Identifier-Of-What's-Been-Contributed-After-That-Past-Time-And-Default-License-of-Future-Contributions: since these might point to different licenses. The snippet construct can possibly express this adequately in some cases but I think reliable identification of a snippet will normally be impractical. Richard On Sun, Jul 17, 2022 at 3:18 PM McCoy Smith <mccoy@...> wrote:
|
|
|