Re: Proposed Update to SPDX License Expression Language
The ‘+’ operator is used to represent the case we see so often:
You may choose to use version <N> of license <L> or any later version of license <L>.
The intent is to provide a way to represent the source file license notice text that often accompanies the GPL and LGPL licenses. GPL-2.0+ is not so much a license, but a license choice (GPL-2.0 OR GPL-3.0). Should there ever be a GPL-4.0 version of the GPL license then it would become a choice between GPL-2.0 or GPL-3.0 or GPL-4.0. Another common use is when one states that a source file or program is under the GPL without specifying a version. In that case, many conclude the license expression to be GPL-1.0+ (because all versions of the GPL state you can pick a version if one is not specified). On a few occasions I have seen some use the “or later version” with other licenses (e.g., EPL-1.0 or a later version) but it is not that common. As was noted previously, the importance of the “+” operator becomes even more apparent when one looks at the many different special exceptions that accompany the GPL license. Some exception notices make reference to the current version only while others state a specific version or later. This was adding complexity to the license list. Making it an operator eliminates complexity.
>> As such, a programatic approach to understanding the nature of the "+" would be complex.
>> Does the "+" refer to anything that is not common ("A" on the examples above), or as inferred,
>> anything after the "-" in the specification, and in that case, only the first "-" as a separator?
In the current SPDX License List (1.19) and the current version of the SPDX spec (1.2) the “+” is nothing more than a valid character used to construct identifier names. That is also true for the character “-“. There is no intended semantics for these characters (in license list 1.19 or spec 1.2). For example, the current “GPL-2.0” identifier could have just as easily been called “GPLv2.0” or “GPL2”. Here even the “2” and “.” have no semantic meaning as well.
Moving forward, in future license lists and specs, the proposal is to remove “+” as a valid character used to craft a license identifier and LicenseRefs, and make it a unary operator that has a specific semantic meaning. This would not change the use of “-“ as a valid character for crafting license identifiers. The “-“character will continue to have no semantic meaning. So the license identifiers GPL-2.0 or LicenseRef-23 are simply aliases for the “GNU Public License Version 2” and a non-standard user defined license respectively. There was no intent in crafting the name where semantics are applied to the “-2.0” characters found in the GPL-2.0 identifier. A tool should not apply any semantics to these characters of the identifier’s name. A tool should simply map the label “GPL-2.0” to “GNU Public License Version 2”.
>> I am also confused with the usage intended. There are 2 concepts that are affected: The license identifier in the License list,
>> and the SPDX License Identifier tag. While similar, they serve different purposes.
There may not be 2 separate concepts. I think the intent is to have just one concept that is consistent with both the license list and spec identifiers. I will try to address your concerns by specifically confirming certain points below.
>> 1. Remove the "+" from the License Identifiers on the Licenst List. (http://spdx.org/licenses/)
>> a. The License Identifier in the SPDX.org maintained tables should not include the "+".
This is consistent with the current proposal.
>> 2. Add the "+" to the License Identifier tag definition specification,
>> a. The SPDX License Identifier tag specification may indicate that the developer has a choice to apply either a specific license as it
>> appears on the licenses list, or all subsequent licenses of the same kind by adding the "+" sign to SPDX filesin
>> the SPDX License Identifier tag after the idString, and before the WITH exceptions.
In the proposal the “+” now behaves like AND and OR in that it is an operator used to construct license expressions. SPDX spec 1.2 does not make it very clear how to construct license expressions. If the proposal is accepted you should expect the “+” to be removed from the definition of idString:
[idString] is a unique string containing letters, numbers, “.”, “-” or “+”.
And more clarification will need to be provided on how to construct license expressions using +, AND, OR, WITH and (). Gary O’Neall recently pointed out that we need to also include an order of precedence. Here is the current proposed ordering:
>> 3. Eliminate the "-" as a possible value of the LicenseRef (see the BSD 4 examples above.
>> Is LicenseRef BSD-4, or just BSD? Will "BSD-2+" a tag, or "BSD+", or could be both?)
As I noted above the “-“ is just a valid character use to construct License Identifiers and LicenseRef identifiers with no intended semantic meaning. The “+” was a valid character in 1.2 but we are proposing to remove it as valid character in 2.0 for both license identifiers and IdStrings.
>> 4. Make the "-[idString]" mandatory, (see Aladdin example above)
>> a. Need to define a default value to be used as -idString, in the absence of one in
>> the specification, although it is likely that all of them have one (Aladdin is v8, as per the web page)
I am not sure I understand this requirement. Are you suggesting that all licenses should have a version number as part of their identifier name?
>> I am a new member of this community, and I do not yet grasp fully the SPDX 1.2 specification, let alone the WIP 2.0.
>> I hope I am not restating the obvious.
I appreciate the level of detail you have reviewed SPDX spec 1.2. Feedback at this level is particularly important to the spec’s evolution and success.
>> As a separate question, I was not able to find in spdx.org a current draft of the 2.0 specification. Is one available already?
It is not yet ready for prime time. It is being drafted and available to those participating in the technical working group. I hope you can participate.
Mark Gisi | Wind River | Senior Intellectual Property Manager
Tel (510) 749-2016 | Fax (510) 749-4552
From: Daniel Companeetz [mailto:dcompane@...]
A comment about the "+" operator.
The standard SPDX Licenst list on the Appendix 1 of the 1.2 Specification, contains entries without the version qualification.
As such, a programatic approach to understanding the nature of the "+" would be complex. Does the "+" refer to anything that is not common ("A" on the examples above), or as inferred, anything after the "-" in the specification, and in that case, only the first "-" as a separator?
The License Identifier (short id) definition from the spec 1.2 states
5.1.4 Data Format: "LicenseRef-"[idString]
[idString] is a unique string containing letters, numbers, “.”, “-” or “+”.
I am also confused with the usage intended. There are 2 concepts that are affected: The license identifier in the License list, and the SPDX License Identifier tag. While similar, they serve different purposes.
As I understand, and just for clarification, the actions being proposed are:
1. Remove the "+" from the License Identifiers on the Licenst List. (http://spdx.org/licenses/)
a. The License Identifier in the SPDX.org maintained tables should not include the "+".
2. Add the "+" to the License Identifier tag definition specification,
a. The SPDX License Identifier tag specification may indicate that the developer has a choice to apply either a specific license as it appears on the licenses list, or all subsequent licenses of the same kind by adding the "+" sign to SPDX filesin the SPDX License Identifier tag after the idString, and before the WITH exceptions.
b. Indicate that the operator only applies to the idString portion of a License Identifier, which may contain a "-" per the specification
c. Add to the proposed specification above that the "+" operator indicates all versions that are comparable as strings bigger or equal than the identifier indicated. You may want to indicate a plain ASCII comparison, or make a reference to the matching Guidelines (https://spdx.org/spdx-license-list/matching-guidelines) for, for example, the purpose of capitalization.
3. Eliminate the "-" as a possible value of the LicenseRef (see the BSD 4 examples above. Is LicenseRef BSD-4, or just BSD? Will "BSD-2+" a tag, or "BSD+", or could be both?)
a. Maybe the license list needs to have separate fields for the LicenseRef and the other components in the table (and one with the complete string?)
4. Make the "-[idString]" mandatory, (see Aladdin example above)
a. Need to define a default value to be used as -idString, in the absence of one in the specification, although it is likely that all of them have one (Aladdin is v8, as per the web page)
I am a new member of this community, and I do not yet grasp fully the SPDX 1.2 specification, let alone the WIP 2.0. I hope I am not restating the obvious.
As a separate question, I was not able to find in spdx.org a current draft of the 2.0 specification. Is one available already?
Please feel free to contact me for questions or clarifications, or if you would like me to attend a meeting to further discuss this message.
On Wed, Apr 9, 2014 at 3:35 AM, Philip Odence <podence@...> wrote:
Editor’s note: Big thanks to Mark Gisi for driving this important effort.
In the wake of a highly productive cross-functional meeting at the Linux Collaboration Summit, the SPDX Legal Team proposes certain changes to the SPDX specification and license list. We are now eliciting comments on the proposal for the next two weeks from the larger SPDX organization which we present below. Please send any comments or questions to the three of us.
- Jilayne Lovejoy [opensource@...]
- Paul Madick [paul.madick@...]
Statement of the problem: The current SPDX Specification and License List have a limited language (i.e. and/or) that is not suitable for expressing complex license situations (for examples of such complexity, see http://wiki.spdx.org/view/FileNoticeExamples ). Additionally, the current SPDX License List currently requires the addition of multiple licenses to the list in order to capture combinations of versions, the “or later” license genre (i.e. GPL-2.0 and GPL-2.0+) and the most common exceptions (i.e. GPL-2.0+-with-autoconf-exception and GPL-3.0-with-autoconf-exception). It should be noted that the SPDX Legal Team identified more than 30 discrete exceptions. Finally, license exceptions (i.e. autoconf-exception) are generally not versioned, and therefore, accounting for changes to the license exceptions when they are part of the canonical license creates opportunity for error and confusion. For additional background please see http://wiki.spdx.org/view/Legal_Team/License_Expression_Review_1
Aim: Improve the ability to describe more license variations utilizing some operator(s) without impact to the current timeline for the next version of the SPDX Specification.
(1) “+” to be added to the SPDX Specification as an operator to denote “or later.” As a result, all current licenses on the SPDX License List with a “Full Name” that includes “or later” will be deprecated. Deprecated means the license entry will technically remain on the list (just moved to the Deprecated section) and list users will be highly discouraged from using them. Deprecated licenses are potentially subject to removal in a future version of the list.
· Example: The license GPL-2.0+ will be deprecated. License entry GPL-2.0 will remain on the SDPX license list and the following is still a valid license expression “GPL-2.0+”. This state’s GPL-2.0 or later because the “+” now plays the role of a unary operator that means ‘the version of this license or a later version’. The following would also be a valid license expression as well: EPL-1.0+ which means EPL-1.0 or a later version. The change of the “+” to an operator will make even more sense once one considers license exceptions (modifiers) in the next section.
(2) “WITH” binary operator (similar to AND/OR) to be added to the SPDX Specification. It denotes that the license identified by the SPDX short name identifier is modified by a well-defined exception on a new “SPDX Modifier List” as described in section (3). As a result, all current licenses that include the term “exception” in the “Full Name” will be deprecated.
· Example: “GPL-2.0-with-bison-exception” will be deprecated but the following expression, which achieves the same result, is now valid: “GPL-2.0 WITH bison-exception”.
(3) The Legal Team will be responsible for creating and maintaining the canonical list of SPDX License List modifiers (i.e. “auto-conf- exception, bison-exception”). It should be noted that licenses that are deprecated will be able to be expressed (more eloquently) with the new changes; hence no license will “fall off” the SPDX License List, nor will this create a compatibility issue. URL references to all deprecated licenses will be maintained for full backward compatibility.
· For example: the following is now a valid expression: “GPL-2.0+ WITH bison-exception” where the “+” and “WITH” operators allow a more flexible expression to be constructed. Prior to this addition this expression was not easy to express.
Longer Term Solution: Add capability for users to define modifier references local to an SPDX document, analogously the way LicRef is used for licenses not on the License List. This will involve adding “ModRef” to the SPDX Specification so the following would be a valid expression: “GPL-2.0 WITH ModRef23.” Where ModRef23 provides the text of a license modifier that does not appear on the SPDX Modifier List. In the initial proposed solution, the way to express a file with a GPL-2.0 license with an exception, that is not included on the SPDX Modifier canonical List, will be by including the full text of the license and the exception as a single LicRef (which is the current state of affairs as well).
Rationale: The proposed solution will provide the SPDX community the ability to express more licenses and license combinations in a more elegant and meaningful (human consumable) way then is currently available under the existing SPDX Specification and License List. To the extent that the SPDX community has come to rely on any of the licenses superseded by the new specifications those licenses will still be available on the SPDX License List (although deprecated). Finally, the initial proposed solution is a major improvement to the current situation and is easily extensible to the even more flexible longer term solution without impacting the current timeline for SPDX 2.0.