Note that this has a GPL3 (not LGPL) license. I’m curious about the rationale for doing that with a template-based data structures library. Why would you want to limit adoption so much?
(Edit: realized after typing this that yes, LGPL wouldn’t change anything for a template library. Question remains though.)
Probably the same reason as with other libraries from the academia. When all free licenses are pretty much equivalent from your point of view, choosing a license is just a bureaucratic formality. The person choosing the license usually doesn't think about the issue too much, and they probably don't have that much experience with licenses anyway.
The currently active fork of SDSL (https://github.com/xxsds/sdsl-lite) has a BSD license. There was supposed to be a team developing the fork, but most people either left the academia or had their research interests move elsewhere. Now there seems to be just a single developer who maintains SDSL as a byproduct of SeqAn.
What surprises me is that I would think the usual goal of an academic project is to make it usable as broadly as possible (indeed, the least restricted licenses originated from universities). And that every institution would have figured out license guidance so it wouldn’t be an accidental choice.
Not sure if this is the case here, but you can change the license of an open source project by unanimous permission of every contributor. (Possibly removing the contributions from those who disagree)
This will not revoke the licenses for copies already in use, but any future users will be subject to whatever license the version they're using uses.
The person who changed the license was one of the main developers of SDSL and one of the coauthors of the papers describing the library. They made plenty of commits to the library, as you can see by looking at the commit histories of individual files. They don't appear on the list of contributors for whatever obscure reason related to the internal workings of GitHub.
Their commits are probably using an email not associated with their GitHub account, which is, as far as I know, how GH associates attribution.
Maybe GH should list contributions not associated to any account by their commit email directly, instead of hiding the contributors, but that's a different matter :P
As it seems that there is just one copyright holder, the obvious answer is that it makes buying commercial licenses more likely option.
For a small company or single person who sells code for a living, GPL is a great license as a first option. People can play with the code and use it for software that is GPL compatible. For everything else you must come up with a coin.
Oh wow, seeing this on the front page makes me nostalgic of my bachelor thesis for the creator of SDSL 10 years ago. I remember getting the question "did your resulsts get merged into SDSL?" during the defense and, unfortunately, I had to say no because they weren't useful in the end.
Glad to see Simon seems to be doing well and the project found some attetion!
If you work with impossibly large but mildly compressible data, this library is like a magic book that gives you all the incantations needed to make your objectives achievable on commodity hardware.
15 comments
[ 3.5 ms ] story [ 36.3 ms ] thread(Edit: realized after typing this that yes, LGPL wouldn’t change anything for a template library. Question remains though.)
1. LGPL doesn't help header code.
2. It not practical to change a license after receiving outside contributions without copyright assignment.
And maybe also 3 they don't have the energy to pursue violators and don't really care.
The currently active fork of SDSL (https://github.com/xxsds/sdsl-lite) has a BSD license. There was supposed to be a team developing the fork, but most people either left the academia or had their research interests move elsewhere. Now there seems to be just a single developer who maintains SDSL as a byproduct of SeqAn.
How is that legally possible? As I understand, the GPL doesn't allow a derivative work to remove the copyleft provisions from the license.
Or maybe I misunderstood your comment?
This will not revoke the licenses for copies already in use, but any future users will be subject to whatever license the version they're using uses.
- the commiter who changed the license has zero commits in the upstream repo
- the commit message has no other context or links to a discussion
Maybe GH should list contributions not associated to any account by their commit email directly, instead of hiding the contributors, but that's a different matter :P
For a small company or single person who sells code for a living, GPL is a great license as a first option. People can play with the code and use it for software that is GPL compatible. For everything else you must come up with a coin.
Glad to see Simon seems to be doing well and the project found some attetion!
https://www.amazon.com/Compact-Data-Structures-Practical-App...
I haven't read the book, so I can't comment on the quality.