[OSM-legal-talk] MAPS.ME combining OSM data and non-OSM data?

[Christoph Hormann]

On Sunday 10 July 2016, Ilya Zverev wrote:
>
> Let's consider another use case. An application that shows OSM map,
> and on top of it shows 1 mln of user points. A users has an option to
> hide the OSM map underneath proprietary points, with a radius of 1
> km. Does in that moment when a user clickes the options, the combined
> map become derivative? Because the application removes parts of OSM
> map based on proprietary data, which means, by your implications,
> that that creates an inseparable references.

I would keep it on the level of combining proprietary data and OSM data 
for the same feature type because this is what you do and this is also 
what is best documented in the guidelines and related discussion.

As i see it you acknowledge that there is such a combination of 
different data sets but since you have a reverse case in comparison to 
the examples given in the guidelines they do not apply and you somehow 
read the license itself to support your use case.

I think this is an interesting viewpoint although i see little chance of 
this becoming a widely accepted interpretation.  It depends on the idea 
that when generating your produced work or publicly using the two data 
sets in combination you have a Collective Database and no Derivative 
Database.  This is going to be really hard to argue since you just 
modified one of the databases you combine for the obvious purpose of 
using it in combination.  Removing hotel POIs from OSM only makes sense 
if you use it in combination with your other data set - the 
de-duplicated OSM part of your alleged Collective Database is therefore 
clearly not an independent database.

If you think through this scenario somewhat further it would essentially 
mean share-alike to be ineffective in de-duplication cases.  Since 
de-duplication is generally only possible in cases where both data sets 
have a roughly comparable quality level (though not necessary the same 
level of completeness) it will hardly ever matter from a practical 
viewpoint which data set you remove duplicates from.  So if one 
direction was possible without share-alike the guidelines would 
essentially be irrelevant because they'd only distinguish between those 
cases where you have to de-duplicate in one direction and those where 
you can combine data sets freely without share-alike.

-- 
Christoph Hormann
http://www.imagico.de/