Seriously wonderful to see this trend. Also really interesting to observe the different approaches to open licenses and attribution: - Overture Maps Foundation is "generally" available under Community Database License Agreement – Permissive v2 (CDLA)[1] and doesn't require attribution although they let you know how to attribute them if you wan to[2]. - Foursquare's open data is available under Apache 2.0[3], which requires attribution for downstream products, which I expect to be pretty unwieldy for a dataset this comprehensive. Software doesn't splinter up into a million different applications the way I expect this data to. - The EU explicitly recommends Creative Commons Public Domain Dedication (CC0) but does provide for an allowance to require attribution under Creative Commons BY 4.0 [4]. Open data is not the same as open source code and I have a hunch that attribution and share-alike licenses are likely to inhibit adoption of open data products (and don't get me started on non-commercial licenses 🙃). Regardless, I'm happy with this trend! I think licenses will be ironed out over time, but curious to hear if people have more informed opinions on this (paging John Wilbanks and Dave Platz!). [1]: https://lnkd.in/guBQ4SUZ [2]: https://lnkd.in/gEshsRMs [3]: https://lnkd.in/gDP2yh-3 [4]: https://lnkd.in/gKWRMRa3
2023 and 2024 have been absolutely fantastic years for open data. Just to mention three fundamental releases: 1. Overturemaps: first non-beta dataset in 2024, now grown to more than >50.000.000 places https://overturemaps.org/ 2. Foursquare: just released >100.000.000 places https://lnkd.in/da6Myb6f 3. EU High-Value Datasets Regulation (2023/138): in force since June, finally forcing EU member states to open all kinds of datasets like parcel data, DOPs, DEMs/DSMs etc. Read through the annex what kind of spatial datasets to expect https://lnkd.in/dGa8EmrS With all three however, the challenge is data harmonization. Peaking at the category systems of Overturemaps (>2000 entries) and Foursquare (>1000) the question remains how to bring OSM and those two together. Image from Foursquare data explorer: https://lnkd.in/d_FiXDFK
Chris Holmes
3w
I'm also curious about the license choice, but more from the standpoint of Apache 2 being a software license, and not an actual data license. I looked into this a bunch in the past and the conclusion I reached after talking to many people is that it's not an appropriate license for data. It is really centered on software and it's unclear how its terms apply to data. The OdBL at least is designed for data. When I looked into things Creative Commons didn't apply to data, and they weren't even interested in it, but I believe the latest CC licenses work very well for data. I do Apache 2 makes the 'intent' clear - attribution only. But I think licenses that are designed for data will help more people feel comfortable with using this dataset.
OSM is the basis of Overture though?
I expect the trend will go on the other direction- with more bespoke requirements- for the reason you state. Licening data is very different than livensing source code, no matter how much people wish they were the same.
Associate General Counsel at Amazon
2wJumping in a bit late on this, but I think existing open source licenses aren't always a good fit for datasets and bespoke licenses will make it harder for open datasets to proliferate. My hope is that "open" licenses for datasets standardize and the industry adopts them so there's less friction to using open datasets. Most open source licenses were not drafted with data in mind. Some open source licenses use terms like "work of authorship" or "derivative work" to grant rights and attach obligations on how the licensed materials can be used, but because these terms have specific meaning under most Copyright legal frameworks, it's not always clear what rights and obligations attach. (What's the work of authorship when there is no author?) This lack of clarity could create friction to adoption. One of the benefits of open source licensing is that a large body of public knowledge exists explaining how these licenses are generally interpreted, allowing the public to generally understand what they mean without having to consult a lawyer. Bespoke licenses don't have these benefits and so the lack of industry practices also creates adoption friction.