[OSM-talk] source=Google

[Christoph Hormann]

On Saturday 18 May 2013, Yohan Boniface wrote:
> On 05/18/2013 03:55 PM, malenki wrote:
> > Dave F. wrote:
> >> IMO Source should be on the object, not on the changeset.
> >
> > +1 (except if there is one changeset for one object (; )
>
> This is not my opinion. Let's take a simple example: a school.
>
> [...]

The problem is currently neither changeset nor object tags are really a 
good solution for true metadata (that is information characterizing the 
data and not the real world object).

Changeset tags have mainly two problems:

- they always apply to the whole changeset so everything you map 
together needs to have the same metadata.  This might seem to be a 
problem primarily for imports but it can also be troublesome in manual 
mapping - imagine mapping something based on satellite images and you 
need to use different images for various parts due to clouds or even 
the common case of supplementing survey data with Bing images.

- many large objects are included in a lot of changesets without 
actually being substantially modified (like moving a single node in a 
500 node way etc.)  This means finding the actual changeset a certain 
geometry originates from to get the metadata information is not so 
easy.

The solution in my opinion would be to have separate metadata tags which 
are reset everytime a substantial change is made to the data they refer 
to unless the user explicitly sets them (either individually or for the 
whole changeset).  Geometry metadata tags for example would be reset 
if:

- in case of a node the node is moved
- in case of a way more than X percent of the nodes are changed (X being 
something like 30)
- in case of a multipolygon more than X percent of the ways are 
added/removed or substantially modified

This would not be fool proof of course (small changes could accumulate 
to a substantial change without being noticed).

Greetings,

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