The idea behind my questions is to differentiate as much as possible tag edits and geometry edits.
The idea is that updating tags in a database is a straightforward job compared to updating geometries that usually needs much more resources. Any improvement in the database update process that can limit these reconstructions will be a real benefit reducing the needed resource to maintain a GIS database updated. In that perspective current diff files are perfect to maintain an API (non GIS) database updated but inappropriate for GIS db. One complete different thing that would also been very interesting to have in even more augmented diffs is the previous version of an object. This would allow to do analysis on edits without the need of a database... and also apply diff backwards. Of course, the side effect will be larger files, but is that a real problem ? Last question (for the moment)... have you looked at .o5c file format to produce near pbf compact diffs ? 2012/8/28 Roland Olbricht <[email protected]>: >> A few questions... > >> > >> When the change on a node or way is just a tag change and does not > >> involve geometry change, are the diff still providing all linked > >> nodes/ways ? > > > > Yes, they do. > > > > Thank you for pointing this out. > > > > I think it would be resonable to not include a way when the only change is a > tag change of the connected node. However, the opposite is may or may not > apply: > > > > Imagine somebody runs a database of all italian restaurants in Paris. He > stores for each way that is the outline of an italian restaurant the center > coordinates. > > > > Now somebody reclassifies a restaurant from whatever to italian, or corrects > a typo like cuisine=italien to cuisine=italian. Then the database maintainer > cannot derive the coordinates of the way because the underlying nodes are > not contained in the augmented diffs. > > > > Of course you could argument that the same applies to relations, and that > the database maintainer could ask for all nodes once he knows that a > relevant restaurant appeared. > > > > So a general question to all: should we keep off the nodes of ways that have > not changed their geometry? > > > > Cheers, > > > > Roland > > -- Christian Quest - OpenStreetMap France - http://openstreetmap.fr/u/cquest _______________________________________________ dev mailing list [email protected] http://lists.openstreetmap.org/listinfo/dev
