[OSM-legal-talk] Licenses for Produced Works under ODbL

Paul Norman penorman at mac.com
Mon Oct 22 21:22:28 BST 2012


> From: Frederik Ramm [mailto:frederik at remote.org]
> Sent: Monday, October 22, 2012 11:53 AM
> To: legal-talk at openstreetmap.org
> Subject: Re: [OSM-legal-talk] Licenses for Produced Works under ODbL
> 
> Another interesting question is how easy the algorithm you specify must
> be. It is clear that the algorithm cannot include "buy some Navteq data
> and then do this", or "buy ArcGIS and then do that" - but what if the
> algorithm includes "run this code, it will take 1000 days", or "make
> sure your machine has at least 1 TB of RAM, then continue as
> follows...".

I see nothing that requires "the method of making the alterations to the
Database (such as an algorithm)" to be easily run or rely on freely
available or open-source software.

So although Navteq is out (being additional contents), I see nothing wrong
with relying on ArcGIS to process the data. For example, suppose someone
wanted to import OSM into an oracle database and then ran some
post-processing oracle-specific scripts on it. They could hand over their
scripts and the code used to import it into the DB but neither would run
as-is unless you also had an oracle license. A copy of the database could be
even less useful, but that's clearly a copy of the entire derivative
database in machine readable form.

I see long-running or computationally resource intensive algorithms coming
up in two ways. 

The first is when whatever you're doing just takes that long to run and
there isn't a faster way. I spent 30 days importing to an apidb, but outside
OSM some databases become truly massive. I could see the release for a
scientific database being: here's the exact code we ran, but it took two
months on our supercomputer cluster. In a case like that, it's meeting the
license. Some data just takes a long time to process.

The second is a more interesting case. You might have a case where you have
two methods of making the alterations, one of which is quick and the other
of which is computationally intensive. My reading of the ODbL is that you
have to provide the one you used because it calls for "_the_ method", not
"_a_ method".

So, back to the examples you gave, provided they used a machine with at
least 1 TB of RAM, they'd be fine releasing a method that relied on having
that much RAM. 




More information about the legal-talk mailing list