[Tagging] demolished buildings, temporal component of data

Ilpo Järvinen ilpo.jarvinen at helsinki.fi
Fri Apr 20 18:07:10 BST 2012


On Fri, 20 Apr 2012, Martin Koppenhoefer wrote:

> In some regions of the world OSM is already in a state where many of
> the map modifications are not due to missing or wrong data, but result
> from actual changes in the real world, e.g. a building gets
> demolished.
> 
> Given that we store not only the actual state of the DB but also
> record all kinds of changes that the mappers apply, I wonder if we
> shouldn't agree on some formal mechanism to distinct the changes where
> the map gets updated to the real world from those where the edit is
> done to correct mapping errors, to increase the level of detail or to
> store them for the first time.
> 
> Since the introduction of API 0.6 we have in theory one powerful tool
> where this detail can already be associated to the edit: the changeset
> comments. The only missing link for effective automated evaluation
> would be an agreement on a formal way of storing information there
> (and quite some discipline in structuring your edits and uploads ;-)
> ). E.g. we could use hashtags to distinguish free text from formal
> comments ( e.g. #demolishion , #new_construction ,etc)

Changeset comments/tags are very problematic because they're not fixable 
once you realize you made a mistake.
 
> An alternative could be, e.g. for a building that was demolished, to
> explicitly "map" this. Given an object tagged with building=yes we
> could change the tag to building=demolished, upload to the server, and
> in a second step delete the object and upload again. The deletion and
> second upload could even be automated easily in the editors, if we
> could agree on something like this.
> 
> As an advantage with the second method you would not need to structure
> your edits and changesets in a special way, I'd expect to get more
> reliable results and less oversight with this approach.
> 
> Is someone already using a scheme for this kind of information?

was: prefixes and keeping geomtery which also helps to prevent somebody 
too eager from redrawing from imagery (which ahs certainly happened 
multiple times around here). :)

-- 
 i.



More information about the Tagging mailing list