warn about possible license conflicts when installing extensions

[felix winkelmann]

In a discussion on chicken-hackers the request came up to give the user a warning if an egg depends on another one that has a viral license, or which causes a license-incompatibility.

[Ivan Raikov]

I disagree with this. It is generally not possible to determine license compatibility because there might be egg license with additional exception clauses. Such a warning would be an obnoxious and unnecessary feature.

[felix winkelmann]

Replying to iraikov:

I disagree with this. It is generally not possible to determine license compatibility because there might be egg license with additional exception clauses. Such a warning would be an obnoxious and unnecessary feature.

Believe me, I haven't got the slightest interest in this whole licensing stuff. But the request came up and this ticket is only to collect implementation suggestions and what to do at all. And I do think that giving salmonella or chicken-install or whoever the option to indicate the dependency of a non-GPLd egg to a GPLd egg is a useful thing.

[Jim Ursetto]

My opinion, for what it's worth. I think that since even lawyers have a difficult time figuring this stuff out, there's no way a computer will. I think the current system is working fine -- which is that salmonella dependency charts include the license for each egg, and we can check this manually. I'd rather keep this "taint check" out of chicken-install, as well as keeping any "viral bit" or out of .meta files. It's a recipe for disaster; at some point we will have endless discussion and disagreement about what form of linking or usage consitutes "viral" -- exactly what is happening on the list now.

I posted a simple script to the list a few months ago that walks the dependency chain of any egg and shows the license. I think Mario took inspiration from this script for the salmonella license checks. Anyway, in my opinion it's easy enough to eyeball the script output or the salmonella output, and this should be enough.

[sjamaan]

I think having a "viral" flag is wrong and silly. It's not generic enough and too obviously hostile against the GPL.

But I don't see why having a simple list of what licenses are known to conflict with eachother to prevent honest mistakes like the one Kon made is so offensive.

Exception clauses just make things more difficult than necessary, effectively creating or forking off another new (and unvetted!) license each time it's done. I think "License X with exception Y and Z" is a hairy enough concept to require a forced override by the user. Something like that is so specific that it should probably be decided on a project-by-project basis anyway.

But what do I know. I just regret we're in the middle of another stupid licensing discussion again :(

[Jim Ursetto]

Kon argued that there was no licensing conflict, and he may have been right. That's why it's not so simple.

Past that you start getting into the weeds (I fully confess I am totally ignoring that thread) so I thought it would be better for our sanity just to punt.

Now, I'll let other people provide actual constructive advice. ;)

[felix winkelmann]

There seems to be no agreement here. Should be close this ticket?