[OSM-legal-talk] ODbL: How obscure/inaccessible can published algorithms be?

[Matt Amos]

On Sun, Dec 13, 2009 at 6:27 PM, Anthony <osm at inbox.org> wrote:
> On Sun, Dec 13, 2009 at 1:08 PM, Matt Amos <zerebubuth at gmail.com> wrote:
>>
>> On Sun, Dec 13, 2009 at 3:37 PM, Anthony <osm at inbox.org> wrote:
>> > On Sun, Dec 13, 2009 at 2:37 AM, 80n <80n80n at gmail.com> wrote:
>> >>
>> >> The example I described above clearly demonstrates that you can't
>> >> differentiate between company A who doesn't use a derived database and
>> >> company B who does.
>> >
>> > What if company C makes a derived database and gives it to company D?
>> > Does
>> > company D have to release the derived database?
>>
>> no. if company D is a subcontractor to company C and no produced works
>> are published. if either company C or D publish produced works from
>> the database, they must make an offer of it. if company D isn't a
>> subcontractor then company C must make an offer of the database.
>
> Okay, so if company C makes derived database and gives it to company D, then
> company D creates tiles with that database, company D has to offer the
> database to anyone who receives the tiles, right?

yes, if D is a subcontractor of C. otherwise both C and D must offer it.

> However, if company D downloads the original database from OSM, then company
> D creates tiles with that database, company D doesn't have to offer the
> database to anyone who receives the tiles, right?

they've almost certainly created a derivative database, for example if
they're using postgis+mapnik, so i'd say they would have to offer that
database.

> Rereading the ODbL, this seems like the most natural way to read it.
>
> Assuming these two points are true, what is considered the original
> database?  Anything on the official (planet.openstreetmap.org) download
> site?  Only databases which were created by OSMF employees?  Only the raw
> on-disk PostgreSQL datastores?  Something else?

technically, it's the on-disk postgresql datastore, plus the server
implementation. the planet is a database dump, and loading that into a
database is creating a derivative.

> If I distribute a tile on March 31, 2010, what exactly do I need to offer?
> The exact portion of the database which is used to create this tile?  If the
> data later changes, I still need to keep the old version in case someone
> takes me up on my offer, right?  Is it enough to keep the full history and
> expect people to look at the timestamps to figure out the state of the
> database at the time of their download?

no. LWG took legal advice on this and it's sufficient to provide the
latest version of the database, or whatever you have which is as close
to the version the user used as possible.

> Can users decline the offer, in which case I can delete the database?  Can I
> give users the option to download the database immediately or to decline the
> offer, so I don't have to keep historical data around indefinitely?

it's not necessary to keep historical data. and you don't have to keep
dumps around either. the offer is pretty much "if you contact me, i'll
give you my database as close as i can to the version you used". if
you practically can't keep the dumps, then that's not a problem.

if you delete all records of the database, then your only options are
to recreate it, or reveal the method used to create it.

>> > Do they have to mention
>> > company C?
>>
>> if D produces works, or further distributes the database or a
>> derivative of it then yes.
>
> What if company C gives them permission not to, or if company C asks them
> not to reveal who they are?

attribution is at the company's option, so if company C doesn't want
to be attributed then D can't mention them. the reverse is also true,
if company C wants to be attributed then D can't remove that
attribution notice. of course, neither C or D can remove the
attribution to OSM, as OSM wants to be attributed.

cheers,

matt