Ghislain Vaillant writes ("Re: Bug#875876: RFS: python-dtcwt/0.12.0-1"):
> FYI, here is the interpretation of the license by the upstream author. I
> asked about it back when I did the initial release, and no issue was
> raised by the FTP team.
>
> https://github.com/rjw57/dtcwt/issues/109
Quoting that:
All Python code and documentation in this repository is BSD 2-clause
licensed. The wavelet co-efficient data is effectively in the public
domain being as it is merely a mathematical "truth" that those are
the wavelet co-efficients.
The original implementation of the DTCWT includes this clause:
[troublesome wording]
Note that the original license uses the language "use the
algorithms" which is unclear. The code in this repo implements the
DT-CWT algorithm and if an algorithm is covered by copyright in a
given juristiction, any use requires referencing the original
author. If algorithms are not copyright-able then use does not. I
have no control over this and I don't want to try to interpret the
original license.
So I think that the situation is perfectly clear. Algorithms are not
covered by copyright (anywhere). The upstream author is just being
over-cautious in leaving in that notice.
I suggest that in the Debian package we c&p that comment from the
github page into the source package, and put that next to the upstream
caveat about the original clause - and then remove the caveat and the
troublesome notice from debian/copyright.
Ian.
