GNU GPL and Open Source Definition
Seth David Schoen
schoen at loyalty.org
Wed Apr 28 07:14:11 UTC 1999
Nick Moffitt writes:
> 28Apr1999 03:03AM (+0000) From [jalami at gcs.bc.ca] jalami [Jeff Alami]
> > AT http://www.gnu.org/copyleft/gpl.html#SEC3, in Article 3, the GPL
> > states that you may sell binaries and executable code at any price
> > you see fit, without corresponding source code. If you do so, you're
> > required to provide source code at a subsequent request. However,
> > unlike what's said in the Open Source Definition, you may also
> > charge for the source code (to a reasonable extent, of course). The
> > Open Source Definition Article 2 states that "there must be a
> > well-publicized means of downloading the source code, without
> > charge, via the Internet."
> You are correct in that this does not meet the letter of the
> OSD. However, the FSF appears to have taken a "genie out of the
> bottle" approach to this issue.
> After all, if I charge $1000 for my CD of source code (perhaps
> the companion to the $5 binary CD that I sold you last week), all you
> need do is take up a collection and then buy it as a group. Then you
> can burn your own CDs and put it up on the Internet for free. Once
> the genie is out of the bottle, it's impossible to put it back in. I
> think that prohibitive pricing is easy to spot, and likely defensible.
> I believe that Bruce Perens did this with a geological survey
> CD, yes?
This point is instructive. US government documents are not copyrighted
and are automatically public domain (with the quasi-exceptions of classified
information and other data "exempted" under the FOIA). Most of the time,
anyone who obtains a copy of a government publication is entirely free to
make and distribute copies (as long as the copies don't claim to be
The Bureau of the Census sells copies of the TIGER/Line dataset (which I'm
not supposed to call "TIGER" because of the registered trademark). They
charge over $1,000 per copy, which is called a "user fee" (the idea being
that government services which benefit particular people could be billed
to those people, rather than subsidized with tax revenue). So the
(relatively few) users of this obscure government service are supposed to
help pay for it by this user fee.
The interesting assumption that the Census people make is that relatively
few people will want to resell or otherwise redistribute a complete set
of usable TIGER/Line data -- otherwise the Bureau wouldn't be able to collect
such a large fee.
This assumption actually has some basis in reality. First of all, who uses
TIGER/Line data? Well, people going off on solar car races, to be sure, and
then authors of mapping and navigation software and services. What do those
people have in common? Well, they tend to be fairly competitive. If Garmin
pays $1,000 for some information, do they presumably
(a) want that information to be available to the whole world, and so send
a copy off to Vicinity, gift-wrapped, with a friendly greeting card
asking that, next time Vicinity staff happen to be near Olathe, Kansas,
they drop off a few CD-Rs to help replenish Garmin's supply; or
(b) hope that Vicinity will have to pay every penny of that $1,000 to get
a copy for themselves?
Since companies that could benefit from TIGER/Line data for proprietary
purposes are in competition with one another, they have no real incentive
to try to save one another money.
But even people who already purchased a copy of TIGER/Line datasets from the
Bureau of the Census to "scratch an itch" aren't _necessarily_ just going to
let the genie out of the bottle. Internet bandwidth remains a somewhat scarce
resource, and six full CDs (i.e. nearly 2^32 bytes) is _a lot_ to let random
And someone who wants to buy TIGER/Line data in order to resell it (for a
profit or at cost) has a fairly large initial capital outlay for an
individual. That basically meant that it took a charitable donation in
order to get this _public domain_ genie out of the bottle!
I think this example proves that even formally public domain information
can still be effectively kept out of the hands of many people who'd like
it. If you have the only copy of some free software, and decide to sell
copies for tens of thousands of dollars, the prospect that someone will want
to buy a copy and then resell it at lower cost is far from certain.
Sure, there's a theory that the price will gradually decline and approach
distribution costs, thanks to successive for-profit resale of the software.
But that scenario supposes that there is a competitive market for obtaining
copies of the software, which is true of Civilization: Call To Power, but
might not be true of TheoretiCAD Super Enterprise Edition Plus.
Richard Stallman was clearly concerned about this issue as it applied to
distribution of source code, and decided to impose a price cap on the
separate distribution of source. Therefore, the GPL specifically does not
take a "genie out of the bottle" approach to source code (although it
seems that Stallman's particular concern was that _nobody_ could afford
the source, not that only a small number of people could afford the
This is really an important issue. Think of how much government
information, for instance, is available on a user-fee basis. (I include
in this all the materials which could be requested through the Freedom of
Information Act or its state counterparts, such as the California Public
Records Act. Under FOIA and its kin, you have to pay a research and
copying cost to obtain the information; this could run into the thousands
of dollars or more, depending on what you ask for. While there are some
fee waivers under FOIA, they don't exist under CPRA.) The "genie out of
the bottle" effect _should_ supposedly mean that all of this information
would be cheaply available to anyone who wanted it. It should all have
been posted on the Internet, right? Or you should be able to buy it on
a CD for $7.95? I mean, here's all of this potentially _very useful_ and
_very interesting_ information in the public domain...
But, of course, there are all sorts of economic reasons that this public
domain information has not actually escaped to public view. Not everyone
can afford to make certain ambitious requests. Many of those who can are
using the information they request for proprietary purposes. Many people
would consider it too difficult to try to redistribute the information
they actually end up obtaining.
Even TIGER/Line data, which is a relatively specific and straightforward
thing, even advertised on a Census web page, goes out of date; it's
updated and reissued periodically by the Bureau of the Census. So this
will require a charitable donation _every two years_ or so, if the free
software community is to stay current.
I mention all of this in order to demonstrate that the copyright and
licensing status of a piece of software or other information is _not_
enough to result in its availability to the general public. There are
many distribution schemes which will tend to keep the current version of
something "off the street", even if the license invites licensees to
share and enjoy.
Depending, again, on how broad the OSI's mandate is to consider _both_
licensing and distribution practices, this sort of thing might become
more than an academic question. What if a software company, for instance,
decided to distribute its new Open Source package under some sort of
TIGER-like terms? "$30,000 for the 6-CD set ... and then give a copy
to whomever you choose." Whoever can raise $30,000 for that purpose is
probably not going to want to give out a lot of copies.
Seth David Schoen <schoen at loyalty.org>
They said look at the light we're giving you, / And the darkness
that we're saving you from. -- Dar Williams, "The Great Unknown"
http://ishmael.geecs.org/~sigma/ (personal) http://www.loyalty.org/ (CAF)
More information about the License-discuss