[OSM-talk] Licence of Facebook's derived road datasets? ODbL?

[Michael Collinson]

I suggest this is "referencing" and, while it does not mention the word, 
is covered in the Legal FAQ 
https://wiki.osmfoundation.org/wiki/Licence/Licence_and_Legal_FAQ#Can_I_use_OSM_data_and_OpenStreetMap-derived_maps_to_verify_my_own_data_without_triggering_share-alike.3F

I think I originally wrote this. Perhaps the LWG would consider if there 
is reasonable consensus and add machine data training as a third example.

Here is a thought experiment to test it:

1) If you notice something interesting in Google Streetview or on an 
in-copyright map and copy it into OpenStreetMap, that is a no-no.  But 
if you go to the location and verify it for yourself, perhaps taking 
your own photos, that is OK. You have used the third-party resources as 
a reference. However, you have then done your own original research and 
based your OpenStreetMap contribution on that.

2) Several years ago, the South African government mapping directorate, 
(who have been very friendly and cooperative with us), wanted to monitor 
OSM for changes, perhaps using machine algorithms. They could then send 
a mapping resources to just those places and remap them. This saves 
enormous amounts of budget in frequently resurveying the entire country 
or large parts of it. Was that OK given that not all their re-survey 
might find its way into open data sets? The LWG at time considered this 
was OK, because of the referencing principle that, while it 
"helped/aided/assisted", it did not involve copying/extracting our data.

3) So, I suggest that it is a logical extension that machine data 
training (and perhaps back testing too?) certainly "helps/aids/assists" 
but does NOT involve  copying our data - then it is referencing rather 
than deriving. As one early thread responder suggested, this is a grey 
area. But my strong feeling is that a liberal rather than restrictive 
interpretation is more helpful to us in growing or map and user base 
than not.

Mike


On 2019-11-15 10:29, stevea wrote:
> I don't know.  I've expressed my opinion(s) on the matter, and believe the LWG should chime in with "an" (the?) answer.
>
> SteveA
> California
>
>> On Nov 14, 2019, at 3:27 PM, Martin Koppenhoefer <dieterdreist at gmail.com> wrote:
>> sent from a phone
>>
>>> On 15. Nov 2019, at 00:19, stevea <steveaOSM at softworkers.com> wrote:
>>>
>>> But the "ultimate test" of "can the new work be made without OSM data?" remains a good one, in my opinion, because then, the author can be told, "well, then, go do so, please, otherwise offer us attribution of some sort" (whether legally required, or not).
>>
>> if you distribute a dataset and say: all roads but not those in OpenStreetMap, isn’t this already attribution? The question is whether you’d want to force them to distribute under ODbL rather than MIT (and maybe what the downstream users have to attribute).
>>
>> Cheers Martin
>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk