Relicense the library to GPL?

[ziotom78]

When I started working on this package, my idea was to provide a fresh rewrite of the algorithms used in the original chealpix library. Halas, this has proved to be a too optimistic solution: in the next years, I have been adding routines by heavily relying on the original chealpix source code, which is issued under a license (GPL) different from the one used here (BSD), slowly converting its C++ routines in Julia.

My main concern when I created Healpix.jl was to provide the Julia community with a package easy to install and supported on all the platforms running Julia; I must admit I did not put too many thoughts on the license to use (I picked BSD because this was the one used by Julia itself).

I am thinking of re-issuing this library under the GPL license: this would be fair to the authors of the chealpix license. However, I would like to know what other people think of this.

[giordano]

My understanding is that translation of a code to another language, having had a look at the original code, is considered derived work, thus GPL requires this to be GPL.

[mreineck]

I'm certainly not complaining :) but are you aware that the central Healpix functionality is now also available under BSD terms? If you are interested, see https://sourceforge.net/projects/healpix/files/

[ziotom78]

@mreineck, thanks for the info, I wasn't aware of that!

However, my primary concern with this Julia package is its availability on all the platforms supported by Julia, which include Windows. (As far as I understand, it is not trivial at all to compile and install the C++ Healpix library under Windows.) I would not include the BSD C library in Healpix.jl, as this would require users to install a C compiler as a dependency. And I fear that the alternative to provide precompiled binary blobs bundling the C library (like FITSIO.jl does) would be too burdensome to maintain.

My idea of implementing everything in Julia was not to bypass the GPL license, but to make the code easier to port to other architectures, without requiring a C++ compiler, while exploiting the capabilities of Julia+LLVM to produce fast code.

[giordano]

And I fear that the alternative to provide precompiled binary blobs bundling the C library (like FITSIO.jl does ) would be too burdensome to maintain.

I can help you with that, as long as the library can be cross-compiled 🙂 What I'm saying, this is a pure Julia implementation 😅