Housenames (aka numbers) are important to do proper geocoding of addresses, e.g. for routing applications. There’s the Karlsruhe Schema which defines one approach how we could do it. As we developed it, we wanted a solution which is easy to map, so the rest of the team anticipated that relations should be avoided. As a consequence, mappers now stuff the full sermon of information onto each individual housename node redundantly:
addr:house_name
addr:country
addr:town
addr:post_code
addr:street
There’s another proposal for a relation which glues the snippets of a street together. Given the fact that there already are 1767 relations of k="type" v="street" in our database, we just could continue to use them:
This method has some advantages over other methods. It’s robust against renaming and even splitting streets. As long as one segment of the original street remains in the relation, it won’t break the housname link. It further reduces redundancy and is not that difficult to understand and map. As I’m pretty sure we will have one relation per street real soon now(tm), I really wonder if others will see the benefit and start using it.
