[Imports] Blog post about imports

[Ilpo Järvinen]

On Thu, 13 Mar 2014, Christian Quest wrote:

> Regarding import related tags, some are really useless, some like unique IDs
> have not been sucessfully used in many cases up to now, but this doesn't
> mean we will never be able to take advantage of them.

I've never really understood why one would _ever_ want to have the foreign 
IDs, unique or not, for things that should remain spatially stable 
(roads, buildings, etc.) because the foreign dataset should be rather 
trivially diffable against older version of itself. That way, the IDs of 
the foreign data are pretty useless to begin with because the diffing 
produces the very cases that would need to be checked regardless of IDs. 
Obviously the previous version(s) need to be preserved though for the 
diffing to be possible.

Besides, the foreign unique IDs usually don't have well defined rules 
anyway, so in case they redo something there would be rather little 
guarantees on how the foreign IDs will be in the end. Which is pretty 
similar results as what we get ourselves with OSM ids when we edit 
by splitting and joining.

For POI type data where re-location is much more realistic compared with 
rather fixed features such as roads and buildings, I can somewhat 
understand that there might be some use for the foreign IDs but even there 
the update process would probably get most out of the foreign data by 
diffing foreign dataset against itself first.


-- 
 i.