Aggregation as derivation [LWN.net]

Aggregation as derivation

Posted Mar 28, 2009 19:33 UTC (Sat) by man_ls (subscriber, #15091)
In reply to: Midway v. Arctic International by butlerm
Parent article: OSBC: Life at the edge of the GPL

Or how about the suggestion that "mere aggregation" creates derivative works?

Yes, quite irrelevant. It doesn't matter because the GPL allows "mere aggregation", so it is not important if the result is a derivative work or not.

Actually the GPL doesn't even care about derivative works, it says that you can distribute the software with or without modifications if you comply with certain conditions -- and "mere aggregation" with proprietary software is allowed. Of course you cannot distribute original or derivative works without a license, and the GPL wouldn't even kick in if you are not doing this distribution; but once you do distribute you have to follow the rules in the GPL (not copyright law).

IANAL (in case you haven't guessed).

(Log in to post comments)

It's so obvious it's not even funny

Posted Mar 28, 2009 20:02 UTC (Sat) by khim (subscriber, #9252) [Link]

Or how about the suggestion that "mere aggregation" creates derivative works?

Actually GPL specifically says that "mere aggregation" is allowed because of course "mere aggregation" produces derived work (such compound works are governed by special rules in many jusrisdictions thus they are not "usual" derived works, but they are derived works nevertheless). Collected stories were inveted hundreds of years ago - and usualy they need specific permissions from all authors. If Joe hates Jane and says that he does not want to see their works under some cover... well - that's it. Compare with GPL which specifically says that authors don't care if unrelated story is placed under same cover.

Of course you cannot distribute original or derivative works without a license, and the GPL wouldn't even kick in if you are not doing this distribution; but once you do distribute you have to follow the rules in the GPL (not copyright law).

That's not exactly true: copyright law can supersede GPL but since default is "you have no right to distribute this period" GPL can force you to do many different things. In many countries you have rights beyond license (for example usually you can change anything in program - even in binary, proprietary blob - if it's the only way to make program work on your hardware), but usually all such rights are limited to private modifications (you can publish info about how program should be modified to work with hirdware like yours but generally you can not distribute modified version without license).

Mere Aggregation

Posted Mar 30, 2009 15:54 UTC (Mon) by butlerm (subscriber, #13312) [Link]

I should clarify what I meant to say here. Of course distribution of a "mere aggregation" on the same medium creates a derivative work in the form of the aggregate or collection. What it doesn't do is make the elements of the whole derivatives of eachother.

So the copyright holders of a software module can control how, where, and when that module is distributed, and in particular what other works it can be inseparably distributed with (e.g. on the same CD). That is well established. References to such collective works can be found in federal copyright law.

However, publishing a collection of stories most certainly does not, in and of itself, make each of the stories derivative works of each other. Suppose you statically linked proprietary code into an open source program, or open source code into a proprietary program. That certainly makes the whole a derivative work, but it certainly does not make, in and of itself, the components derivatives of eachother. That was SCO's rather ridiculous theory - i.e. that e.g. Linux JFS was a derivative work of a proprietary Unix because it some point in its history its predecessors had been linked to proprietary code.

So my point is (barring some bizarre judicial invention) a derivative work (logically at least) has to be "substantially similar" in some inclusive way to the work it is "based upon". "Access and substantial similarity" sounds like a good rule of thumb to me, as long as an exception is made for purely functional interfaces at the boundaries.