fix Apache License for join.py

[kanzhang]

Not sure if this is the right thing to do, just in case it was overlooked.

[AmplabJenkins]

Can one of the admins verify this patch?

[pwendell]

@kanzhang ah I see - I think this probably should remain as-is if it's code from another project. I thought you meant there was no license on this file.

[kanzhang]

@pwendell I'm no expert on this, do as you see fit (I'm still a little surprised - I thought you give up your copyright when you upload your patch to Apache).

[srowen]

No Apache does not take over copyright in contributions, IIUC. Contributors own their contributions. In this case, was this code 'contributed' by the owner (Douban?) to begin with? or just added to the code base? If the latter, it can still be reused, if the license is compatible (and it looks so) but if the author didn't specifically agree to license differently, you can't decide that for them. Leave the license unless it's clear that this was contributed by the copyright owner under AL2.

[pwendell]

@kanzhang I'm no expert either unfortunately :P. I do think in general we ask (maybe require) authors to sign over copyright to Apache when they contribute.

I think this code might be a somewhat abnormal case, I noticed @JoshRosen actually added this code (from UCB) in the original pyspark commit. I suspect the reason is that he was borrowing source code from an existing library and pulled in the corresponding license. In the case where pull in source code from other libraries I believe this is the correct thing to do.

[kanzhang]

Close as invalid.