[OSM-talk] Why the BSD vs GPL debate is irrelevant to OSM

[Frederik Ramm]

Hi,

    this would normally be a posting for legal-talk or so, but since the 
topic has been brought up here on talk, I'll respond here.

I wasn't about to revive the debate. I prefer PD but I'll support the
move to ODbL. I'm not trying to convince everyone that PD is better.

However, I feel that the PD cause has been gravely misrepresented
by people claiming (a) that PD vs ODbL is like BSD vs GPL, and (b) that
BSD is dead. I'm not going into the details of (b) which is obviously
wrong; I want to explain why (a) is wrong, which is less obvious.

Many people think: "Share-alike has proven it worth in other areas, so 
it must be good." - It is certainly true that share-alike works in some 
areas. Computer programs and works of art come to mind. Let us first 
look at computer programs: Since the GPL (GNU Public License) was 
invented, the Free Software Movement has had a huge success developing 
an enormous amount of high quality software. Much of this software is 
GPL or under some other kind of share-alike license.

If you look at the software stack on a modern computer, you see the 
kernel, above it libraries and then the application programs. The most 
popular free kernel is the Linux kernel which is under the GPL, but 
libraries and applications used on top of a Linux kernel don't fall 
under the GPL restrictions. (They may be GPL licensed but they are not 
GPL licensed *because* they run on a GPL kernel.) Some licenses are 
LGPL, allowing non-GPL software to use them. So there is ample precedent 
that it is good for basic components not to dictate the licensing of 
stuff that is built on top of them.

Share-alike is always a two-sided sword; you make life more difficult 
for some of your users (e.g. those who want to build a modified Linux 
kernel), in order to make life easier for others (the Linux users who 
know they will get access to any kernel modifications). This share-alike 
element has to be balanced in order to work; if you make *too* many 
demands of those building a modified kernel, they'll just not do it, 
giving a sub-optimal result for the whole community.

The overwhelming use of a Linux kernel is in a "collective" situation - 
you use the kernel as-is, together with other bits and pieces, to make 
up a working computer system. The majority of people don't want to take 
the kernel apart an modify and extend it, and thus are completely 
unaffected by the share-alike component of the license - even though 
they of course do lots of things on their computer that they couldn't do 
without the Linux kernel.

Another very similar example is the Creative Commons movement. There are 
many photos and pieces of music and even films out there that come under 
a Creative Commons Share Alike license and the model is commonly 
considered a success. It is great that I can use a photo I find on 
flickr in a presentation. Its great that I can use a piece of music in a 
podcast without problems. And my whole presentation or podcast doesn't 
have to come under the same license as the original work just because of 
that. Again, as with the Linux kernel and the stuff that people build on 
it, I can choose any license they want for the presentation or podcast. 
Sure, if I use some CC-BY-SA image and change it a little bit, I can't 
change the license. But the overwhelmingly predominant use of Creative 
Commons licensed work is not in this way, but by including the work in 
some larger work where the "collective work" instead of the "derivative 
work" rule gives me the right to do anything I want.

Now back to OpenStreetMap. A very typical use of OSM data at the moment 
is to create maps from it and then build some kind of application on top 
of that. The map is still CC-BY-SA, the application on top of it isn't. 
Here as in the case of a photo used in a presentation, the license for 
some part doesn't infect the whole. But there are many many more uses of 
this data. Taking the OSM data and mixing it with other data from 
outside is where the interesting and new applications lie, and what you 
will likely be doing for routing or geocoding or location-based services 
or anything else of interest. This is something different. This doesn't 
happen that much in the world of art or the world of programming. The 
intimate mixing of sources, or thorough processing of one source, makes 
a lot of sense with the data. Ask yourself the question: How much more 
likely is someone to mix data from several geodata sources than to 
re-mix a few CC-BY-SA licensed music tracks?

This means that any share-alike provision we slap on OSM data has a much 
more direct influence on the potential uses of the data than a 
share-alike provision on software on on creative works has.

ODbL tries to reduce this problem by exempting "produced works" from the 
share-alike effect, and this is a good thing, but still there will be 
many use cases adversely affected by the remaining share-alike for data.

To recap; the share-alike component in geodata will affect vastly more 
usage scenarios than a share-alike component in software or works of art 
ever will. The share-alike provision of GPL, or CC-BY-SA, affects only a 
fraction of the users of the so licensed work. In contrast, the 
share-alike provision of ODbL (or CC-BY-SA if it could validly be used 
for geodata) affects the majority of users of the licensed work, and 
thus is a much stronger restriction.

This doesn't automatically lead to the conclusion that there should be 
no share-alike component; but it makes clear why it is foolish to claim 
that we're leading "just another GPL vs BSD" debate.

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"