[linux-elitists] ultra 20 m2

Rick Moen rick@linuxmafia.com
Thu Jul 12 13:13:39 PDT 2007

Quoting Steven Critchfield (le@drunkenlogic.com):

> A derivative work is one that _INCORPORATES_ portions of other code. 

Actually, that incorporates substantial amounts of copyrightable
expressive elements.  Such copying can be either literal or non-literal:
See, for USA jurisdictions, the decision in Micro Star v. FormGen, Inc.,
154 F.3d 1107 9th Cir. 1998, where a competitor's game program was ruled
derivative even though it didn't have a single line of code in common
with the original; the copy incorporated audiovideo display data, a
protected creative work, from the original, thereby incorporating the
"story" (not the story idea, but its various expressive elements) of the
original game.

> This is why some people consider the GPL to be a viral license.
> Specifically if you can't stay at arms distance from it,....

...which means nothing more than avoiding using someone else's
copyrighted creative works without permission....

> you are likely going to be required to release under the terms of the
> GPL.

Rubbish.  Infamous misinformation, at that.

Let's say I was working, some years back, at a Linux firm when one of
these issues came up.  I could probably name the firm without harm
because it can't seem to make up its mind what it name is, but it's a
bad habit, so I won't.

Let's say an e-mail request came in, one morning, for matching source
code access per the terms of GPLv2 clause 3b from a pleasant-sounding
guy at Mountain View Data, who noted that the firm had published a
binary RPM of an ndmp backup utility coded from a programmer in, say,
New Hampshire.  He noted that the firm's version, which it published as
part of a Red Hat variant distro for clusters, seemed to have
improvements, and he wanted to see them.

This was referred to me as the firm's unofficial licensing guy.  The
request sounded sensible to me, and so I inquired with Software
Engineering about source -- and they panicked.  It turned out that some
failure of communication between Professional Services and Softwar
Engineering had permitted some code from two of the firm's major
business partners, which was proprietary and shared under NDA, had been
added to the GPL ndmp by one department, and then released to the public 
by the other department without double-checking code provenance.[1]  I
sent a summary of the situation to Jay S., the chief company counsel, 
waited half an hour, and went to knock on his door.

Now, Jay is a very sharp fellow, and I was always proud that Jay
considered me relatively clueful for a non-lawyer, so he would often 
let me work things out rather than hand me explanations on a platter.

RM:  Let me guess:  We're not going to provide source code to the
Mountain View Data guy.

JS:  No, we aren't.

RM:  But we _are_ obligated to do so under the terms of GPLv2 cluse 3b 
that we accepted when we made derivative works and redistributed, right?  

JS:  Yes.

RM:  Hold on.  Let me work this out....  Failing to provide the source
violates the New Hampshire guy's copyright, which is a tort, because
we've been carrying out actions permitted only if we meet that
obligation....  This is going to be one of those situations where we are
forced by circumstances to choose between two torts, and so are picking
the one that's harmless, right?

JS:  Right.

RM:  OK, complying with the request would mean violating our contracts
with [names of the two heavy-hitter business partners], and being in 
breach of contract with them would be very bad, subjecting us to
possible serious tort litigation as well as souring important
relationships and ruining our reputation generally.  So, very bad
outcomes if we do that.

On the other hand, let's say that we cease offering the modified ndmp
package to all outside parties right now.  We've been violating the New
Hampshire guy's copyright in a very minor way up until now, but (1) he
almost certainly didn't know about it, and didn't complain, (2) he's
probably suffered about zero demonstrable economic loss aka "actual
damages", which is all (other than enjoining further infringement)
that's even potentially available unless you've registered your
copyright with the Library of Congress either prior to infringement or
within three months of publication, (3) almost no open source coder
bothers to register (which allows collection of statutory damages plus
attorney's fees) because it takes $35 and some paperwork.

The New Hampshire guy would have standing to sue, but isn't motivated,
isn't annoyed at us, and couldn't collect anything -- and we're right
about to fix the only thing that could annoy him.  The Mountain View guy
completely lacks standing concerning the ndmp code (not being the
affected copyright owner), and therefore couldn't even potentially sue

JS:  Right.

RM:  So, we're going to reply back to the Mountain View Data guy saying,
"We're really sorry for the inconvenience, but the modified ndmp code
was made available in binary RPM form accidentally, and should not have
been.  Accordingly, we've removed it from the ftp site."   And thus we'll
have fixed the problem _and_ done the right thing.

JS:  You got it.

Another case that should be more familiar:  NeXT, Inc. started
distributing binary-only releases of modified gcc, hacked to include
ObjC (and, IIRC, C++?) support.  FSF asked politely for the source to
their changes; Jobs and company initially stated that this was contrary
to their company policy.

What happened next is frequently misrepresented, in recountings.  It is
-not- true that NeXT, Inc. was "required" to release its source
modifications to gcc under GPLv2.  It had been committing the tort of
copyright violation, but tort law _nowhere_ has the power to compel 
violators to release their changes to the public.  That's not how it

If Jobs had wished, Next, Inc. could have said something like "Oh, we
misunderstood, but wish to do the right thing, and are therefore ceasing
effective immediately to distribute our modified gcc.  We apologise for 
any problems this causes."

What would FSF's recourse have been?  Absent gcc copyright registration, 
it could have sued for an injunction ordering cessation of infringement
(pointless, as it would have already ceased), and for actual damages,
which would have been nil.  With copyright registration, it could have 
done the same, but sought statutory damages plus attorney's fees.
Statutory damages are in a range subject to the judge's discretion, 
but, if defendant credibly claims to have been "not aware and had no
reason to believe" they were infringing copyright, damages would
typically be $200 per work.[2]  So, FSF would have about zero motivation
to sue, even if they hated NeXT, Inc.'s guts and wanted to punish it.

My point:  NeXT, Inc. _elected_ to give back source code, after the
obligatory initial period when they behaved like jerks, not because they
had to (which they absolutely did not), but rather because it was
greatly to the company's advantage to be able to give/sell customers
compilers with enhanced language support.

So, Steven:  Please do not continue to spread that ridiculous and
erroneous assertion about infringers supposedly being "required" to
release changes.  It's done more than enough damage, over the years.

[1] http://linuxmafia.com/~rick/lexicon.html#stepwise-disaster

[2] http://en.wikipedia.org/wiki/Statutory_damages_for_copyright_infringement

More information about the linux-elitists mailing list