Viral permissiveness

[Steve Thomas]

Donovan:

I almost completely agree with you.

> A person can have copyright on 1000 characters of source code, but they don't have copyright on each individual letter taken separately because a single letter is not original.

Of course, but the larger a substring common to two works, the more
judges would tend to conclude that the works have a common origin that
they are 'based upon', or that one is 'based upon' the other.
Naturally so - the other outcome would be increasingly improbable on
the basis of this information alone. However, authenticated provenance
data might show that one of the substrings is a concatenation of
strings of differing origins, all prior to the other work.

I'm not suggesting the _sufficiency_ of such "provenance" data. I
doubt that the question of whether or not a work is 'based upon'
another is even formalisable in a logic. If it were, it surely could
not be computationally decidable. No, we'll need judges and expert
opinion for that matter a while yet... Rather, there are situations
where provenance data (if it were available!) would be preferable to
the diffs I mentioned in the prior post.




Generally then: Where might consideration of code-provenance be relevant here?

Let's provisionally define a "reciprocative" license for a project as:
  A copyleft license for the project's source code that, upon
distribution of modifications, requires that modifications to original
works licensed to the project are licensed back on the same terms to
the original licensors.

Inherent then, in the problem of crafting an effective 'reciprocative'
license, is the situation wherein a project file contains a plurality
of copyrighted works under a plurality of licenses. As dtemeles has
pointed out:

> 17 U.S.C. § 103(b) provides:

> The copyright in a compilation or derivative work extends only to the material contributed by the author of such work, as distinguished from the preexisting material employed in the work, and does not imply any exclusive right in the preexisting material. The copyright in such work is independent of, and does not affect or enlarge the scope, duration, ownership, or subsistence of, any copyright protection in the preexisting material.


EXAMPLE SETUP

There are four distinct projects producing source code: A, B, C and D.
A-code and B-code are both distributed under non-copyleft licenses.
B-terms have distribution provisions that implicitly prevent
distribution of copies of B-code under A-terms.
C-code is distributed under a 'reciprocative' license as provisionally
defined above.
D is a licensee of C-code and so also distributes under C-terms.
The C-license notice pertaining to the file includes the declaration:

"This collective work is Copyright (C)2009 by C-Project. Individual
portions may be copyright by individual contributors, and are included
in this collective work with permission of the copyright owners."

If I understand [1] correctly, there is an issue, general to all
open-source licenses, over whether or not a judge would uphold the
assertion of collective-work status.

Also, the notice is a bit coy, isn't it? "My license applies to all of
the Code except where it doesn't, where You might get sued by someone
else if you infringe some other terms." All of the examples cited in
the definition of 'collective work' [2] would usually have clear
boundaries - not so here.

Let's continue with the examples anyhow.


EXAMPLE 1

This example just involves A, B and C.

F1 is a file, created by C, comprising a copyrightable modification of
some A-code and some B-code.
Moreover, F1 contains a fragment of B-code that one or more judges
would rule is substantial enough for B-copyright, not C-copyright, to
subsist in.

In this example, C cannot distribute F1 to A under A-terms without
infringing B-terms: reciprocation is impossible. Practically, C could
_try_ to remove the B-code from F1 producing F1a containing the A-code
and F1b containing the B-code, and distribute F1a under A-terms and
F1b under B-terms. This may not always be possible, of course.

General observation from this example:
  (Ia) 'reciprocative' licensors must avoid any immiscibilities of
external licenses with their own license for the collective work, in
common with other open-source licensors.
  (Ib) Furthermore, unlike most other open-source licensors,
'reciprocative' licensors must avoid, per file, any immiscibilities
between the file's external licenses.

Alternatively, I suppose, one could cripple the 'reciprocative'
license to at most one external source per file :(


EXAMPLE 2

F2 is a file, created by C, comprising a copyrightable modification of
some A-code but no B-code.
There is no F1 in this example: C distributes, enclosing a notice that
modifications of F2 must be reciprocated to A.
D receives a copy of F2 under C-terms.
D creates F3, a copyrightable modification of F2 that includes B-code.
Similar to the previous example, F3 contains a fragment of B-code that
one or more judges would rule is substantial enough for B-copyright,
not D-copyright, to subsist in.
However D either:
 (D1) mistakenly does not believe he has included sufficient B-code
for this to be the case
or
 (D2) is unfamiliar with the extra (Ib)-immiscibility issue given above.
D then distributes, complying with C-terms by, amongst other things,
licensing F3 to A on A-terms.
D has thus infringed B-copyright.

I take it B can sue D. Could D claim 'innocent infringement'? Suppose
there are many such infringements like the one in this example. Could
B sue C? Could a judge object somehow to the C-license on the basis of
the "nuisance" of its downstream infringements?


EXAMPLE 3

This is like example 2, but with A and B reversed.

F4 is a file, created by C, comprising a copyrightable modification of
some B-code but no A-code.
Moreover, F4 contains a fragment of B-code that one or more judges
would rule is substantial enough for B-copyright, not C-copyright, to
subsist in.
C distributes, enclosing a notice that modifications of F4 must be
reciprocated to B.
D receives a copy of F4 under C-terms.
D creates F5, a copyrightable modification of F4 and some A-code that
still contains enough B-code for B-copyright to subsist in.
D distributes F5 to A on A-terms, infringing B-terms.

Again, suppose there are many such infringements. Could B sue C? Could
a judge object to the C-license?

In this example, C could have reduced the likelihood of downstream
infringements by including additional information with F4 to suggest
what fragments are of B-provenance. The inclusion of revision history
seems onerous and the "provenance" data mentioned before generally
infeasible. However, a copy of the original B-code should be easy for
C to include - with the reciprocation notice, perhaps - and
distributional provisions preserving that inclusion could be
incorporated into the C-terms.

Maybe I'm wrong, but I think I'm getting a clearer idea of what I want
and beginning to ask sensible questions about it. Time to find
pro-bono help? Any suggestions anyone has in this regard would be
warmly welcomed ;)

Steve

[1] http://www.catb.org/~esr/Licensing-HOWTO.html
[2] http://www.copyright.gov/title17/92chap1.html#101

-- 
My GnuPG key ID is 0x82314996
Preferred keyserver: hkp://subkeys.pgp.net