Am 07.11.2017 um 21:25 schrieb Jonathan Masur:
...
J'ai donc utilisé le système qui était recommandé par Wambacher : Les anciennes frontières communales sont marquées "boundary=historic" et "admin_level=8" avec la date de la fusion des communes sous le tag "end_date". Les anciennes communes sont des relations avec "admin_level=10". Dans tout les cas les dates de début et de fin de communes sont ajoutées au tags, sauf lorsqu'une commune déjà existante s’agrandit sans changer de nom (comme par exemple Servion qui annexe Les Cullayes - dans ce cas "Servion" n'a pas de date de début et de fin).
Ceci à l'avantage d'expliquer clairement la situation - la frontière est historique, mais les "quartiers" de la nouvelle commune existent encore aujourd'hui - par exemple il se peut que la représentation au conseil communal doive être proportionnelle avec les anciennes communes, dans ce cas les anciens territoires sont utilisés.
First a clarification, Wambacher recommendations are based on the situation in Germany, and they are not directly applicable to Switzerland in any case (I would consider a number of his arguments rather hand wavy even for Germany, but that is a different story).
To make a sanity check if my position on this matter was completely off the rocker, I've spent some time researching what the "amtliche Vermessung/mensuration officielle" contains in such situations, and checked this in two cantons that have their information on line. In all cases I looked at the plot information is associated with the (new) municipality and the addresses with the postal city (PLZ6). Matter of fact the AV for the Canton Zürich actually requires you to explicitly switch to a search on the Swissnames dataset to find place names that are neither existing municipalities or postal cities, the AV dataset simply doesn't contain them.
That raises the question how does Swisstopo model such places in the Swissnames dataset (note the dataset is not available on terms that are OSM compatible) as this seems to be the major source for such place names? They draw a rough polygon around the built up area of such places and attach the name to that (so likely the equivalent would be landuse residential+industrial+commercial in OSM), this seems to be compatible with information from Stefan Keller that indicates that the place signs (the on the ground indication that there is actually a place there) on streets are placed roughly at the intersection of the building zones with the roads
In summary there is no indication at all that the municipality borders of no longer existing municipalities are maintained by anybody nor that they model something useful that corresponds at least to a data model or similar that anybody else is maintaining. Given that even in such removed municipality there are often multiple places that are currently only mapped as nodes it would seem to be more useful to agree on a place mapping model that works both for geocoding and rendering and implement that, than adding non-current information to OSM that should be in a different database.
Simon