[OSM-talk] [OSM-legal-talk] Are we strict enough with imports ?

[Felix Hartmann]

On 12.02.2010 12:01, Oliver Kuehn (skobbler) wrote:
> Hi,
>
> Another helpful feature could be an "expiry date". Each import can not be
> valid longer than e.g. 12 months. After this period the dataset needs update
> and receives a new validity data. Otherwise the data will become inactive.
>
> Regards,
> Oliver
>    
It would be nice to be able to tell if imports are used or not. Sadly 
this is not the case. If we removed untouched objects from the Austrian 
plan.at import, then we would surely break even more, because people 
connected other streets to the imported data, without correcting the 
imports....

In general I think data that is easily recordable/traceable shouldn't be 
imported. So streets and their like should have very low priority. In 
the US I think the Tiger Import is at least partly responsible for the 
low interest (it's more fun to enter something new, than to correct old 
stuff). Austria had before the imports actually a on par or better 
coverage than Germany. Then with plan.at imports virtually 80% of all 
roads were inside OSM - but with very low quality. 15 month later 
Austria really lacks in quantity and quality compared to Germany, and 
still IMHO around 30-40% of the Imports are more or less incorrect. I 
think it will take another year or two to recover the damage and get 
down to 3-5% uncorrected import data.

If however we had high-resolution orthophotos cleaning up the import 
would be largely over and only very few bits and pieces would be left 
over and the import probably by now considered as a success.

On the other hand data that is very hard to source, like maybe exact 
postal codes, drains, small rivers, detailed data about landuse (as 
imported in France and Latvia) that without  - or even with - good 
resolution orthophotos cannot be mapped, lower quality could be accepted 
because we stand a hard chance of ever getting it otherwise.

I think the rule, if it is easily mapable (no matter the effort and 
probability of it being done), then we should be a lot stricter and make 
sure that the quality of the import is at least as good as good mapping 
practice allows (Tiger Data is IMHO not good enough quality for 
example). If however it is not easily mapable then imports can be 
introduced.