Abstract Wikipedia Licensing Discussion

Abstract Wikipedia Licensing Discussion

Abstract Wikipedia is a really long time length mission with a unbiased of mixing converse material from unique and present Wikimedia projects to permit contributors to invent and lend a hand Wikipedia articles self sustaining of language. Abstract Wikipedia will in all probability be built using contributor-created utility functions and converse material from other Wikimedia projects, basically Wikidata, to generate text in just a few languages.

Wikifunctions, adore other Wikimedia projects, will depend heavily on contributor-submitted converse material. Wikifunctions can even depend heavily on contributor-driven decisionmaking around all of the projects’ policy pattern. Therefore we must peaceable express together which licenses to make exhaust of for the ingredients of Abstract Wikipedia.

That you just might maybe maybe be taught extra about Abstract Wikipedia and Wikifunctions at Abstract Wikipedia.

The Abstract Wikipedia team has requested the Wikimedia Foundation’s Correct department for an conception on which licenses might maybe maybe be an acceptable option for every and each layer within Wikifunctions, and what other upright dangers must peaceable be thought about in picking a license. This file is in step with their suggestion and guidance.

This web page first describes what these rather a few ingredients are. Then it describes which licenses invent sense for which part. That is adopted by a dialogue of the rather a few alternatives and their interactions with each and each other. The web page is closed by statements of particular particular person opinions.

Overview and interrogate[edit]

(Up-to-the-minute: 2021-12-03)

All contributions to Wikifunctions and the wider Abstract Wikipedia projects will in all probability be published below free licenses. Textual converse material on Wikifunctions will in all probability be published below CC BY-SA, characteristic signatures and other structured converse material below CC-0. We want to express whether to publish Abstract Drawl for Abstract Wikipedia below CC BY-SA or below CC 0, and whether to publish code implementations below Apache or GPL.

We might maybe maybe maybe beget to snatch whether which licenses you might maybe maybe beget, and likewise whether you beget other suggestions, feedback, concerns, and so on. Please pronounce on the talkpage.

We thought to summarise the dialogue and opinions one day around December 16, to head away the draft summary and draft resolution up except December 20, and finalise the resolution factual after or for the length of the office hour, assuming the feedback is definite.


A crucial step in direction of Abstract Wikipedia is Wikifunctions, a wiki of utility functions that might maybe maybe moreover be combined with items from Wikidata. Wikifunctions will in all probability be a utility web web hosting platform, allowing customers to jot down and earn code from their web browser. Wikifunctions will permit customers to publish and earn functions that might maybe maybe maybe incorporate converse material from other Wikimedia projects.

For the sake of licensing, we focus on about converse material at four stages:

  • Operate Signatures: The definition of a characteristic, i.e. its title, number and kinds of input arguments, and the bear of the output;
  • Operate Implementation: The code that’s referred to as and completed within Wikifunctions, which might maybe moreover contain contributor-submitted code and libraries available on the platform;
  • Abstract Drawl: The summary representation of text or text fragments, which is mainly a neighborhood of undoubtedly well-behaved characteristic calls to bear an output; and
  • Output Drawl: The text produced thru calling Operate Implementations on Abstract Drawl, and most ceaselessly pulling in Wikidata Drawl.

An example of every and each is supplied in the proportion below this.

The following image sketches out the rather a few ingredients of Abstract Wikipedia relevant for the dialogue. We checklist the rather a few ingredients in the following text.

Architecture of how Wikifunctions will be used to generate text

Wikifunctions will consist of characteristic signatures, characteristic implementations, and other objects. These other objects might maybe maybe moreover be of numerous kinds, and the space of kinds is extensible. Objects will beget documentation.

Operate signatures are adore APIs, they provide the title of the characteristic and arguments moreover to the types of the arguments and of the final end result. Implementations are the source code of the functions – they explain the pc tricks on how to flip the arguments into an reply. Documentation might maybe maybe moreover be made for any signature, implementation, and another object in Wikifunctions. Assorted objects in Wikifunctions will in all probability be testers, particular particular person strings, lists, kinds, summary converse material, and so on.

One of the considerable considerable functions will in all probability be functions that earn summary converse material and generate output converse material (most ceaselessly natural language text) out of it. One of the considerable considerable functions will entry data in Wikidata or other areas and generate natural language text out of it. The functions might maybe maybe moreover exhaust lexicographic data from Wikidata in remark to generate the text.

The generated text shall be integrated into each and each Wikipedia to beget data gaps.

Instance of the ingredients[edit]

Screen: These examples are extremely simplified, and explain example values in English which would technically be Wikidata QIDs for internationalization (examples).

Given a constructor Superlative with the keys self-discipline, tremendous, class, and pickle constraint, we are able to beget the following summary converse material:

  self-discipline: Jupiter,
  tremendous: tidy,
  class: planet,
  pickle constraint: Solar Plan)

In Wikifunctions, we would beget the following characteristic signature:

generate text(superlative, language) : text

I.e. a characteristic that takes a superlative object (as given in the summary converse material above) and a language (corresponding to English) and returns a text.

That is also then a probable characteristic implementation in Python:

def generate_text(superlative, language):
  if language == English:
    self-discipline = get_label(superlative.self-discipline, language)
    adjective = superlative_form(superlative.tremendous, language)
    class = singular_form(superlative.class, language)
    location_clause = make_location_clause(superlative.location_constraint, language)
    text = ‘ ‘.be part of(self-discipline, ‘is the’, adjective, class, location_clause)
    return text.capitalize_first(text)
  if language == Hausa:

The utility of the characteristic to the summary converse material would end result in the following output converse material:

(in English) Jupiter is the largest planet in the Solar Plan.

(in Croatian) Jupiter je najveći planet u Sunčevom sustavu.

This newsletter can then be shown by a language model of Wikipedia in remark to produce a identical outdated baseline of facts about Jupiter.

Correct prognosis and copyright standards by bear of converse material[edit]

The Wikimedia motion has adopted a overall licensing policy that favors free custom licenses and delivery source utility licenses. The Ingenious Commons Zero (CC0) deed presents a waiver that objectives to guarantee converse material is in the final public domain (or a jurisdiction’s local equal to the final public domain) all the way in which thru the enviornment. The Ingenious Commons Attribution-ShareAlike (CC BY-SA) license enables people to make exhaust of converse material by complying with minimal attribution necessities and guaranteeing that subsequent adaptations are released below the phrases of a adore minded license. At ideal, for utility, Wikimedia follows the checklist of authorized licenses from the Open Source Initiative (OSI).


Details themselves are most ceaselessly no longer protectable below copyright legislation. In the US, the Supreme Court docket described this precept in Feist Publications Inc. v. Rural Phone Service Co.: “Potentially the most main axiom of copyright legislation is that ‘[n]o author might maybe maybe moreover copyright his tips or the info he narrates.’” That is a identical outdated part of copyright legislation that applies no topic whether converse material is explicitly released below a license.

The Wikimedia projects beget adopted CC0 for projects which shall be designed to earn identical outdated info. As an instance, this entails items or properties on Wikidata, or structured data on Wikimedia Commons. CC0 presents a identical outdated, world waiver of copyright to verify that unbiased converse material is free to be frequent with out restriction. This makes it more straightforward for converse material to be found, reused, and cited in other places. This license might maybe maybe moreover moreover incorporate unprotectable info which shall be copied from sources below other licenses, corresponding to text from Wikipedia that’s otherwise available below CC BY-SA.


Utility is in overall protectable below copyright legislation, though no longer all ingredients of utility must peaceable be treated the the same. Wikimedia has taken the say that the group and identical outdated characteristic of APIs are no longer copyrightable as a topic of US legislation.

For copyrightable utility, the Foundation adopts an delivery source utility license to permit it to be freely reused below minimal prerequisites. Wikimedia’s guideline on freedom and delivery source presents that the Foundation must peaceable liberate all of the code it creates below an acceptable delivery source license. In areas the place Wikimedia supports particular person-created utility, corresponding to Wikimedia Cloud products and services, it also requires that the utility be released below an delivery source license. There are a broad diversity of licenses that qualify as delivery source, so to make clear this requirement Wikimedia in overall turns to the checklist of OSI-authorized licenses for utility.

For MediaWiki, the important utility license is the copyleft GNU Overall Public License (model 2.0 or later), and most extensions, skins and PHP libraries shall be found below that license. For another projects, Wikimedia uses a extra permissive license, such because the MIT License or Apache License (model 2.0).

Text or media converse material[edit]

Under the Wikimedia converse material licensing policy, Wikimedia projects might maybe maybe moreover host converse material that’s available below a Free Drawl License, in the final public domain (corresponding to expired copyright or otherwise uncopyrightable converse material), or unbiased exhaust justification (in certain runt situations). For most Wikimedia projects, including Wikipedia this implies that text and media is available below CC BY-SA (model 3.0).

Recommendation from the WMF upright department[edit]

Wikimedia must peaceable put the license for every and each bear of converse material thru certain and simple insurance policies for Wikifunctions. Wikimedia must peaceable invent documentation that explains the licenses for every and each bear of converse material, including the clarification for selecting this license. For utility in explicit, there must peaceable be guidance about handiest importing third-event code below an acceptable license. The particular person interface must peaceable contain a license grant that’s appropriate for every and each bear of converse material.

Operate Signatures[edit]

Recommendation: Operate signatures must peaceable be CC0.

Operate signatures must peaceable be light of identical outdated ingredients (e.g., a list of parameters) which shall be in step with underlying performance. The proper converse material of the characteristic signatures is unlikely to be protectable by copyright, or the place it’s far protectable, it’ll be frequent below unbiased exhaust in the US. CC0 is an acceptable license for data that might maybe maybe moreover no longer be eligible for copyright in any appreciate. Use of CC0 avoids creating confusion or misleading reusers to agree with there are copyright limitations the place none exist. Furthermore, CC0 will guarantee most interoperability with the other delivery source or free custom licenses.

Operate Implementation[edit]

Recommendation: Operate implementation must peaceable be below the Apache License.

Operate implementation must peaceable be below an OSI-authorized license. If the Wikifunctions team and the neighborhood needs to limit this to one license, then the Apache License would provide an perfect stage of permissive flexibility.

Furthermore, Wikifunctions might maybe maybe moreover permit other permissive licenses which shall be adore minded with the Apache License, such because the MIT License or (3-clause) BSD License. Allowing an additional space of license alternatives might maybe maybe moreover permit customers to import extra converse material from other third-event sources. Nonetheless, it would also require creating additional utility necessities, corresponding to a particular person interface to rob a license and point out the appropriate license notices. Wikifunctions might maybe maybe moreover rob a single license, for the sake of simplicity, for the length of the initial delivery and then lend a hand in tips adding enhance for just a few licenses later in step with need.

Abstract Drawl[edit]

Recommendation: Abstract converse material must peaceable be licensed below CC BY-SA or CC0.

Abstract converse material shall be released below a CC BY-SA license or another equally permissive license that matches the mission’s objectives and meets Wikimedia’s licensing policy. Wikimedia has considerable latitude in picking the finest license for Abstract Drawl.

Selecting CC BY-SA might maybe maybe be a customary replacement, and would provide the well-behaved thing about consistency with Wikipedia and most other Wikimedia projects. It will in all probability permit customers to repeat and incorporate Wikipedia converse material into Abstract Drawl in some system. Nonetheless, it would also require preserving an edit historical previous or some equal contribution historical previous mechanism for Abstract Drawl, to permit people to produce attribution to the converse material’s checklist of authors.

Alternatively, Abstract Drawl will in all probability be released below the extra permissive CC0 phrases. This might maybe maybe permit the utility to circumvent the attribution necessities, but limit customers’ potential to repeat or incorporate any protectable portions of Wikipedia articles or other sources.

Output Drawl[edit]

Recommendation: Output Drawl must peaceable be licensed below CC BY-SA or CC0.

Since Output Drawl is generated thru utility that mixes just a few data sources, there shall be questions on whether the following product is copyrightable in any appreciate. In 2019, the US Copyright Issue of job requested comments about how converse material created by AI algorithms or processes must peaceable be handled below copyright legislation. In the Wikimedia Foundation’s submission in response, we defined that AI algorithms must peaceable be treated adore every other utility instrument and that the instrument’s particular person must peaceable be thought concerning the copyright holder. Following the the same precept, Wikimedia might maybe maybe moreover lend a hand in tips Output Drawl as a piece of creativity by the authors of the Abstract Drawl. It will in all probability therefore in all probability be handiest if the output converse material is licensed repeatedly with present Wikimedia projects, and can also be mentioned by the communities as to which of the licenses at mask in exhaust might maybe maybe be preferred.

Suggestions by the enchancment team[edit]

The enchancment team recommends to beget a study the solutions by Correct, that are (to summarize): to rob CC0 because the license for Operate Signatures; Apache for Operate Implementations (and to delivery with a single license, and handiest after we search the necessity for just a few licenses to expand Wikifunctions to enhance just a few licenses); to rob both CC0 or CC BY-SA for the Abstract Drawl and the Output Drawl.

For documentation and other textual converse material of Wikifunctions we are able to rob CC BY-SA in remark to keep compatibility with most other Wikimedia projects relating to textual converse material. For other objects in Wikifunctions moreover implementations, we are able to lend a hand it in step with the replacement for Operate Signatures.

The enchancment team additional recommends to rob CC BY-SA for Abstract Drawl and Output Drawl. Whereas one might maybe maybe moreover argue that Abstract Drawl is extra the same to the structured data of Wikidata than to the natural language text of Wikipedia, we deem that there are a assortment of elements that invent Abstract Drawl sufficiently the same to text:

  1. Editors beget a undoubtedly gorgeous-grained selection of which info are being displayed and that are no longer. In Wikidata we strive for completeness over cautious selection.
  2. Editors beget a undoubtedly gorgeous-grained control of the remark the info are being displayed in, constituting fable ingredients, that are no longer available in Wikidata.
  3. We inquire of the natural language skills to permit editors to proper to a couple stage of emphasis and selection of wording.

All of these level in direction of Abstract Drawl being extra the same to text than to a assortment of info, and therefore we counsel that we notice the the same license that we exhaust for the text in Wikipedia, which is CC BY-SA. Alternatively, one might maybe maybe moreover argue that by striking Abstract Drawl below CC0 we delivery the living for a greater amount of doable reuse in purposes we are able to no longer even agree with but, by no system tips properly express the upright framework for. CC0 most absolutely enables essentially the most freedom in the reuse of the Abstract Drawl.

Request for input[edit]

We might maybe maybe maybe beget to ask the neighborhood to focus on about these solutions and expectantly to search out consensus all the way in which thru the licensing resolution. The target is to lend a hand the dialogue delivery for about four weeks, and, if the biggest, to expand and restructure it. In case this looks to be inadequate to reach a consensus, we might maybe maybe moreover restructure the licensing dialogue to focal level fully on Wikifunctions for now, and then notice up with a dialogue about Abstract Wikipedia.

To manual the license replacement, it’ll be pleasant to lend a hand in tips and focus on about the following questions:

  1. What are the long-time length objectives of the projects, and the way in which might maybe a copyright license enhance these objectives?
  2. Must the people fascinated with creating Abstract Drawl receive credit ranking?
  3. How pleasant is it to keep consistency and compatibility with the licenses on Wikipedia?

(Up-to-the-minute: 2021-12-03)

The explicit inquiries to the neighborhood are the following two: must peaceable we exhaust CC BY-SA or CC 0 for Abstract Drawl for Abstract Wikipedia; and might maybe maybe moreover we exhaust Apache or GPL for code implementations.

Even whereas you deem both of the alternatives are gorgeous, it might maybe maybe maybe maybe be mountainous to head looking out your insist expressed explicitly, in remark to earn a greater figuring out of what the neighborhood tends in direction of.

Feedback are welcome in any language.

Be part of the pack! Be part of 8000+ others registered customers, and earn chat, invent groups, put up updates and invent chums all the way in which thru the enviornment!



Hey! look, i give tutorials to all my users and i help them!