Copyright issues

[kpcyrd]

While preparing hyphenation for debian I noticed there are some issues with the patterns/ folder, most notably:

• Files with custom licenses (eg. https://github.com/tapeinosyne/hyphenation/blob/master/patterns/hyph-en-us.lic.txt)
• Files with non-free licenses (eg. https://github.com/tapeinosyne/hyphenation/blob/master/patterns/hyph-en-gb.lic.txt)
• Files with no license at all (eg. https://github.com/tapeinosyne/hyphenation/blob/master/patterns/hyph-cop.lic.txt)

I haven't checked all of them, but I would recommend deleting all files that are not available with a https://wiki.debian.org/DFSGLicenses license.

Thanks!

[tapeinosyne]

Even though the changes required to remove pattern files from hyphenation would be minor, I should like to mention that those files are already distributed on Debian as a component of TeX Live – or any other TeX distribution with Unicode support, for that matter. (I appreciate that TeX packages may predate the DFSG license policy or be otherwise exempt for reasons that do not apply here.)

Beyond that, I assume that your recommendation is to delete incompatible files not only from the crate itself, as packaged by cargo and distributed through crates.io, but from this source repository as well?

[sylvestre]

Indeed, I wasn't aware of that:
https://sources.debian.org/src/context/2018.04.04.20180416-1/texmf-dist/tex/context/patterns/common/lang-gb.rme/?hl=11#L11
https://sources.debian.org/src/libfizmo/0.7.15-2/src/locales/en_GB/patterns.txt/#L14

I reported: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=912556
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=912557

[silwol]

Any update on this issue? I'm not too deep into these specific copyright topics, but the follow-ups on https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=912557 appear to me as if this could be solved?

[mojca]

"Upstream" @reutenauer fixed the British patterns and uploaded them to CTAN (hyphenation/tex-hyphen@c76c797). If it's not yet fixed here, it's just a matter of syncing, I guess.

I was not actively aware of Coptic, but I'm 100% sure this is easy to fix since the author is really responsive.

The US patterns are original Knuth's patterns (which are apparently acceptable, else the whole TeX Live could be kicked out) + a number of hyphenation exceptions that were later written by someone else whose file we are now shipping. If anyone wants to get Knuth's approval for licence change, feel free to write him (snail mail) and share his response. (I won't be the one asking him.)

If you want to point out other files, please send a list, but please also consult with Norbert Preining.

ConTeXt (mentioned in the Debian tracker) will automatically be updated at some point (the problem is that one cannot easily update to the latest version of ConTeXt without breaking compatibility with the binaries shipped with TeX Live).

[reutenauer]

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=912557 is not an issue but an incorrect reading of the (admittedly convoluted) licence text. In addition, the original author has in the mean time agreed to the MIT licence.

As for the Coptic patterns, their author has given us permission to license them under the LPPL.

[llucax]

I'm also hitting some red flags while doing a fossa.com scan for license over a project that has this crate as a deep dependency. I got 2 red flags:

• File: patterns/hyph-lv.lic.txt is released under LGPL v2.1 or GPL v2-or-later.
• File: patterns/hyph-hu.ext.lic.txt is released under MPL/LGPL/GPL (it doesn't say which version exactly).

But this project is said to be released under Apache-2.0 or MIT (maybe for the second file MPL is compatible, but I'm pretty sure LGPL/GPL is not).

FOSSA offers a free service which is good enough for simple cases, maybe it would be a good idea to scan this project since it's including files from several diverse origins...

[llucax]

File: patterns/hyph-id.lic.txt seems to also be GPLv2-only.

[tapeinosyne]

@llucax, please see prior discussion on the matter at #28 and mgeisler/textwrap#174. As stated previously, those files are part of the hyphenation crate, but are not included in build artifacts.

[llucax]

Ah, thanks! I searched over all issues but never thought that an issue with that title could be related to licensing information. Thanks!

[reutenauer]

Speaking as the maintainer of the upstream package, the situation is definitely not simple, as the individual pattern files have been contributed over the course of several decades, using a wide variety of licences, sometimes more than one for the same file (by request from different projects); and in a few cases the original authors are dead and hence do not respond to requests to relicense their files.

That said, the many licences are all free enough for Debian (we follow the Debian Free Software Guidelines), and the patterns are standalone data files – they’re not libraries – so that as long as your project’s conditions are not too restrictive, you should be able to use them.

[mgeisler]

Perhaps a good approach would be to move the pattern files to a new hyphenation-patterns repository. This way the Rust code here could be fully isolated and everybody should agree that it's licensed under a free license. Put differently, the goal would be to ensure that ~/.cargo/registry/src/github.com-*/hyphenation-*/patterns no longer exists, even if nothing from that directory is pulled into the final binary.

It's already easy to load language files at runtime, e.g., see my interactive example program. Applications using the hyphenation crate can then download and bundle the patterns they feel comfortable with.

If people want to embed the pattern files into their applications, then this functionality could perhaps we extracted into the hypothetical hyphenation-patterns repository.