This blog post is part of our Global Open Data Index (GODI) blog series. Firstly, it discusses what open licensing is and why it is crucial for opening up data. Afterward, it outlines the most urgent issues around open licensing as identified in the latest edition of the Global Open Data Index and concludes with 10 recommendations how open data advocates can unlock this data. The blog post was jointly written by Danny Lämmerhirt and Freyja van den Boom.
Open data must be reusable by anyone and users need the right to access and use data freely, for any purpose. But legal conditions often block the effective use of data.
Whoever wants to use existing data needs to know whether they have the right to do so. Researchers cannot use others’ data if they are unsure whether they would be violating intellectual property rights. For example, a developer wanting to locate multinational companies in different countries and visualize their paid taxes can’t do so unless they can find how this business information is licensed. Having clear and open licenses attached to the data, which allow for use with the least restrictions possible, are necessary to make this happen.
Yet, open licenses still have a long way to go. The Global Open Data Index (GODI) 2016/17 shows that only a small portion of government data can be used without legal restrictions. This blog post discusses the status of ‘legal’ openness. We start by explaining what open licenses are and discussing GODI’s most recent findings around open licensing. And we conclude by offering policy- and decisionmakers practical recommendations to improve open licensing.
What is an open license?
As the Open Definition states, data is legally open “if the legal conditions under which data is provided allow for free use”. For a license to be an open license it must comply with the conditions set out under the Open Definition 2.1. These legal conditions include specific requirements on use, non-discrimination, redistribution, modification, and no charge.
Why do we need open licenses?
Data may fall under copyright protection.
Copyright grants the author of an original work exclusive rights over that work. If you want to use a work under copyright protection you need to have permission.
There are exceptions and limitations to copyright when permission is not needed for example when the data is in the ‘public domain’ it is not or no longer protected by copyright, or when your use is permitted under an exception.
Be aware that some countries also allow legal protection for databases which limit what use can be made of the data and the database. It is important to check what the national requirements are, as they may differ.
Because some types of data (papers, images) can fall under the scope of copyright protection we need data licensing. Data licensing helps solve problems in practice including not knowing whether the data is indeed copyright protected and how to get permission. Governments should therefore clearly state if their data is in the public domain or when the data falls under the scope of copyright protection what the license is.
- When data is public domain it is recommended to use the CC0 Public Domain license for clarity.
- When the data falls under the scope of copyright it is recommended to use an existing Open license such as CC-BY to improve interoperability.
Using Creative Commons or Open Data Commons licenses is best practice. Many governments already apply one of the Creative Commons licenses (see this wiki). Some governments have chosen however to write their own licenses or formulate ‘terms of use’ which grant use rights similar to widely acknowledged open licenses. This is problematic from the perspective of the user because of interoperability. The proliferation of ever more open government licenses has been criticized for a long time. By creating their own versions, governments may add unnecessary information for users, cause incompatibility and significantly reduce reusability of data. Creative Commons licenses are designed to reduce these problems by clearly communicating use rights and to make the sharing and reuse of works possible.
The state of open licensing in 2017
Initial results from the GODI 2016/17 show roughly that only 38 percent of the eligible datasets were openly licensed (this value may change slightly after the final publication on June 15).
The other licenses include many use restrictions including use limitations to non-commercial purposes, restrictions on reuse and/or modifications of the data.
Where data is openly licensed, best practices are hardly ever followed
In the majority of cases, our research team found governments apply general terms of use instead of specific licenses for the data. Open government licenses and Creative Commons licenses were seldom used. As outlined above, this is problematic. Using customized licenses or terms of use may impose additional requirements such as:
- Require specific attribution statements desired by the publisher
- Add clauses that make it unclear how data can be reused and modified.
- Adapt licenses to local legislation
Throughout our assessment, we encountered unnecessary or ambivalent clauses, which in turn may cause legal concerns, especially when people consider to use data commercially. Sometimes we came across redundant clauses that cause more confusion than clarity. For example clauses may forbid to use data in an unlawful way (see also the discussion here).
Standard open licenses are intended to reduce legal ambiguity and enable everyone to understand use rights. Yet many licenses and terms contain unclear clauses or are not obvious to what data they refer to. This can, for instance, mean that governments restrict the use of substantial parts of a database (and only allow the use of insignificant parts of it). We recommend that governments give clear examples which use cases are acceptable and which ones are not.
Licenses do not make clear enough to what data they apply. Data should include a link to the license, but this is not commonly done. For instance, in Mexico, we found out that procurement information available via Compranet, the procurement platform for the Federal Government, was openly licensed, but the website does not state this clearly. Mexico hosts the same procurement data on datos.gob.mx and applies an open license to this data. As a government official told us, the procurement data is therefore openly licensed, regardless where it is hosted. But again this is not clear to the user who may find this data on a different website. Therefore we recommend to always have the data accompanied with a link to the license. We also recommend to have a license notice attached or ‘in’ the data too. And to keep the links updated to avoid ‘link rot’.
The absence of links between data and legal terms makes an assessment of open licenses impossible
Users may need to consult legal texts and see if the rights granted to comply with the open definition. Problems arise if there is not a clear explanation or translation available what specific licenses entail for the end user. One problem is that users need to translate the text and when the text is not in a machine-readable format they cannot use translation services. Our experience shows that it was a significant source of error in our assessment. If open data experts struggle to assess public domain status, this problem is even exacerbated for open data users. Assessing public domain status requires substantial knowledge of copyright – something the use of open licenses explicitly wants to avoid.
Copyright notices on websites can confuse users. In several cases, submitters and reviewers were unable to find any terms or conditions. In the absence of any other legal terms, submitters sometimes referred to copyright notices that they found in website footers. These copyright details, however, do not necessarily refer to the actual data. Often they are simply a standard copyright notice referring to the website.
Recommendations for data publishers
Based on our finding we prepared 10 recommendations that policymakers and other government officials should take into account:
- Does the data and/or dataset fall under the scope of IP protection? Often government data does not fall under copyright protection and should not be presented as such. Governments should be aware and clear about the scope of intellectual property (IP) protection.
- Use standardized open licenses. Open licenses are easily understandable and should be the first choice. The Open Definition provides conformant licenses that are interoperable with one another.
- In some cases, governments might want to use a customized open government license. These should be as open as possible with the least restrictions necessary and compatible (see point 2). To guarantee a license is compatible, the best practice is to submit the license for approval under the Open Definition.
- Exactly pinpoint within the license what data it refers to and provide a timestamp when the data has been provided.
- Clearly, publish open licensing details next to the data. The license should be clearly attached to the data and be both human and machine-readable. It also helps to have a license notice ‘in’ the data.
- Maintain the links to licenses so that users can access license terms at all times.
- Highlight the license version and provide context how data can be used.
- Whenever possible, avoid restrictive clauses that are not included in standard licenses.
- Re-evaluate the web design and avoid confusing and contradictory copyright notices in website footers, as well as disclaimers and terms of use.
- When government data is in the public domain by default, make clear to end users what that means for them.
Danny Lämmerhirt works on the politics of data, sociology of quantification, metrics and policy, data ethnography, collaborative data, data governance, as well as data activism. You can follow his work on Twitter at @danlammerhirt. He was research coordinator at Open Knowledge Foundation.
