Subject: on software freedom and patents
From: "Stephen J. Turnbull" <stephen@xemacs.org>
Date: Sat, 02 Dec 2006 14:39:18 +0900

Thomas Lord writes:

 > What is the relationship between patents and free software?

 > A program is free software if the general public is permitted to do
 > four things with the source code:

Abbreviating (hopefully without loss of clarity):

 >         1) Make copies
 >         2) Study and Modify the program
 >         3) Share the program or modified versions
 >         4) Use the program by running it

 > A patent, meanwhile, is a grant of the exclusive rights to

 >         1) Make
 >         2) Sell
 >         3) Use

 > an invention.

 > A conflict between software freedoms and patents arises only under a
 > very specific condition[...].

 > The four software freedoms can be compared to patent law one by one:

 > * Sharing Publicly Licensed Software Can Not Infringe a Patent

I will assume without loss of generality that sharing implies copying,
not transfer.  See copying, below.

 > * Modifying Publicly Licensed Software (usually) Can not Infringe a Patent
 > 
 >   Quite simply, patent law makes no pretense of restricting the
 >   public right to manipulate the information in the text of a
 >   patent claim,

I think this is false.  If you have a program

(defun use-the-patent-claim ()
  "This program marshals no data and applies a patented device to it."
  (embodiment-of-patent-claim)
)

(defun embodiment-of-patent-claim ()
  ;; I don't have a claim at the moment; let's pretend you can patent `nil'.
  nil)

then you may modify the first defun, but modifying the second one
infringes the exclusive right to make.  (Note that the choice of an
interpreted language makes this clear; however, even for a compiled
language, at some stage somebody will have to infringe the right to
make in order to use the modified program.)

 > * Copying Publicly Licensed Software Infringes Only if Running it Does
 > 
 >   Copying is only an infringing act of "making" if the copying
 >   naturally leads to "use".   Without "use", hypothetical infringement
 >   can not be considered meaningful (or requiring remedy).

Sorry, that's not the way the law works; that's the whole problem.
Consider that courts have interpreted research into patented methods,
and mere publication of modified algorithms, as infringement.

There's a reason why "use" and "making" are separately granted rights.
That is to ensure that mere making, regardless of use or even
usability, is a violation.  (Why the legislature chose to do that is
not in the scope of this post.)

 > * Using Publicly Licensed Software Can Not Infringe
 > 
 >   A patent is, by definition, an assignment of the exclusive
 >   rights to an invention to an inventor.
 > 
 >   The public's right to run a publicly licensed software program is,
 >   by definition, not an exclusive right: it is a freedom for all.

That's false.  The motivation for a public license is to implement a
freedom, but it is nevertheless a license.  That fact that the license
is specified so as to apply to unidentified persons who may not yet
have been born or incorporated doesn't change the fact that persons
are individually granted licenses.

This is the can of worms known as the implicit patent license.  It is
reasonable to assume that when a vendor grants a license to *use*
software, that grant includes all necessary rights owned by the
vendor, and presumably some obligation to reveal other rights that
would be infringed by use if the vendor knows of them.

However, there is no reason why a vendor cannot license patented
software under copyright, but withhold the patent license necessary to
sell.  This is in fact S.O.P, more commonly known as a "sample
implementation of a patent claim".  This would not be a free software
license, but it would be a public license.  Furthermore, you could
imagine that a sufficiently enlightened court would permit the
publication and distribution of a derived program as text as long as
it was not run.  This program could be released under a free software
license, since the author licenses all of his own rights in
conformance to the four freedoms.  But it would not be free software,
since there are necessary rights that the author doesn't own.

Note that I'm not contesting your claim that there is an implicit
patent license.  However, that license derives from the explicit
license to run and the general prohibition of fraud, not from the
public nature of the license.

 > * Selling Publicly Licensed Software Can Not Infringe
 > 
 >   Yet these two kinds of selling are not in conflict.   The right to
 >   sell publicly licensed software is not an exclusive right --
 >   everyone shares it equally.

See above.

It might be nice if patents worked as you propose, but they don't.

It's possible that *in practice* things *could* work out that way, at
least for some free software licenses.  This is the claim that the FSF
makes vis-a-vis the GPL v.2.  However, this is based on the implicit
patent license doctrine, which for several reasons is dubious (not
least that the GPL disclaims all applicability to running the program,
but also including the wording of grants of the rights to make and
sell, which are based in the way we think about copyright law, and
therefore are rather unclear about the scope of grants of patent
license).