Hi, I spend a several days investigating several free software, content, database licenses that could probably suit my objectives, and I believe that the ODbL is the closest to what I have in mind. However, I suspect that my end-users would find it difficult to comply with section 4.3, since my end-users might seek to utilize their Works Produced in real-time within a persistent virtual space. It would be rather annoying if the license were to force them to carry a sign or wear a watermark.
I’m wondering if it would be fine to create a license of my own using the ODbL text as a base, toning down section 4.3 in order to permit for less visible means of providing attribution (e.g. through metadata). Mozilla and the GNU project permit others to craft their own licenses based on their existing licenses, so I’m wondering if the Open Knowledge Foundation permits the same.
Creating new licenses is generally discouraged because it causes data licensed under those terms to be incompatible for reuse with existing licenses. Data users will often need to combine data from different sources, and that might not be legally possible depending on the licenses of each dataset.
Nowadays the ODbL is considered outdated. For licensing new data, if you’d like to exempt data users from any obligation to do attribution (and that is advisable in order to make reuse easier), it’s best to use Creative Commons Zero 1.0 Universal. That license is newer and has been crafted also with data in mind.
The main issue with the ODbL 1.0 is that it is abandonware (it could be argue that it was never anything else). Besides a handful of bloopers, yes, the attribution requirements suffer from a lack of vision and in particular if taken very literally create conflict in practical day to day use.
That said, the OKFN is probably the wrong organisation to turn to for any kind of revision, even non-controversial, no-brainer revisions to the data licences have not been moved forward, see the inability to fix a clear mistake in ODC-by.
In case it wasn’t obvious, CC0 naturally doesn’t do any of the things that somebody leaning towards using the ODbL would want.
And just what would somebody leaning towards using the ODbL want? Attribution requirement? Indeed, CC0 does not demand that.
Requiring attribution in data use is discourage because, while it may be a good practice to cite the sources, it may be impractical in some projects that combine a great number of data sources, you might not have a practical way to attribute all of them and fulfill the requirement. When attributing is a legal requirement, then the dataset simply may not be used in such projects.
If someone, all things considered, would absolutely not forego with the attribution requirement, then they’d better use CC-BY 4.0, as it is also designed with data in mind. I see no possible use case that would justify using ODbL nowadays.
CC-BY 4.0 does not do what you think it does, it is, put very nicely, a “share alike light” licence, definitely not appropriate for somebody that just wants an attribution requirement.
The ODbL is a share alike licence with relaxed terms for non-database derivatives and would only be appropriate for somebody that wants share alike and not just attribution.
CC-BY 4.0 is definitely not “share alike light”, as there is no clause in it requiring that derivative works to be shared with the same licence. For someone who wants both share alike and attribution there is CC-BY-SA 4.0.
The CC 4.0 family of licences and CC-0 have all been written with data use and database laws in mind. So no, I do not recommend using ODbL for any possible use case. It may have been useful at the time it was created (Creative Commons did not address data usage back then), but since 2015 there has been no longer a reason to use it.
Of course, people are free to choose whatever licence they want. However it is important to keep in mind that choosing a licence with attribution requirement or share alike will definitely limit the possibilities of data use and reuse, which is contrary to what one would want when they mean “open data”.
Now I remember reading that discussion back in the day, and it is just one more reason for not choosing to use ODbL. If someone is concerned with allowing downstream derivative works of the data that use with DRM, naming that as “open data” is the misleading part, as you can’t possibly be simultaneously “open” and use DRM.
If the ODbL 1.0 does indeed allow that, it should be removed from the list of licences considered to be “open”. Since the Open Definition is up for review pretty soon, I think this is a point that should be rectified.
The ODbL requires a parallel, non-DRM distribution in such cases and would be totally fine with that. Naturally CC0 would promptly be declared non-open if such a restriction was put in place.