[darcs-users] darcs patch: relicense Darcs.Commands.GZCRCs as BSD3

[Stephen J. Turnbull]

Gwern Branwen writes:

 > Yes, he can. But you're also right: confusingly, what he would've done
 > in such a case is *dual*-licensing:

That's not correct usage.  *Dual licensing* is when a rights holder
offers two different licenses for the same content.  Here we have
different licenses for different parts of the content, because
different rights holders are involved.

 > his changes are BSD,

Caveat: Not if they contain material or structure copied from the GPL
version.  Then he may not distribute his changes under BSD without
additional, explicit permission.

By "structure" I mean that, eg, if the original is expressed as a tail
recursion, and he changes the names of the functions and parameters so
that no original text remains but the algorithm is the same, he's
copied the structure and the original copyright still applies.  If he
instead changes to a loop plus stack structure (as well as changing
the names of functions and parameters), then he is very likely OK
(even though these are computationally equivalent, they are
*expressed* differently).

 > the original is GPL, and the result is BSD+GPL, but note, not just
 > either - since BSD adds no restrictions compared to GPL,

This makes sense if you interpret "BSD" and "GPL" as names for sets of
licensing terms, and "+" as union.  Note that in general it is
possible for the terms in the union to conflict, in which case no
copying or redistribution is legally allowed.  Interpreting "+" as a
logical disjunction is not appropriate; you can't pick and choose
unless a given rightholder explicitly offers multiple licenses.

 > we can essentially simplify it to GPL. This is akin to a Wikipedia
 > editor putting his edits and articles under CC-BY, in addition to
 > Wikipedia's standard GFDL/CC-BY-SA; the legal distinction is there,
 > but in practice, it doesn't mean anything except for new works or
 > works where the entire history (=the whole chain of derivative
 > works) has been relicensed.

In the case of a collective work, it certainly does have meaning.  And
what is a program but a collection of functions?  While not every
function will be original enough to be independent of other
copyrights, or to be itself copyrightable, some subset will be, and I
am certain that it is that set of functions that Ganesh is aiming at.

 > (To continue the Wikipedia example: what are you going to do with an
 > edit which is both BY and BY-SA? You can pick which license you want
 > to use the edit under

No, you cannot.  You must satisfy all the licenses that apply.  If the
new text is a pure insertion, then you can copy it from the patch
under the author's license, but in copying from the derived work you
must satisfy the derived work's license.  So the effective license
must be a superset of the terms of any one author's license (and thus
be weakly more restrictive).

 > But it's valuable if that edit is creating a 100 page article
 > single-handedly, since now you have an entire article in 2
 > flavors.)

Even a single sentence (one line function) can be sufficiently
valuable.  However, it's mostly likely to be worth dealing with at the
level of a functional module or higher (say a subsection in text).

 > If copyright were a proposed type system, it'd be rejected for being
 > logically incoherent. :)

Actually, not.  It's really not that hard to understand how copyright
works, not by comparison with Hindley-Milner ... unless your self-
interest clouds your understanding.  Lawrence Rosen _Open Source
Licensing_ is the best source I've come across.