Address consensus in Croatian settlements

Communities Hrvatska (Croatia)
1

While in communication with locals within Croatia, I have learnt that what I suspected to be street names for addresses is technically wrong, since the streets do not have names.

This mainly applies to smaller settlements from anywhere of 1-500 address nodes and depending on the admin level.

Thus would like to discuss what schema should be developed for such circumstances before modifying the existing nodes and ways.

Streets

For streets would add the noname=yes tag.

Address

For building addresses

addr:housenumber=*
addr:locality=*
addr:postcode=*
addr:city=*
addr:country=*

With addr:locality key being the preliminary choice before possibly specializing to addr:hamlet, addr:village, addr:town, etc. The benefit of the former is that it will persist even if the settlement size changes, thus removing the need to update those tags.

Alternate approach

Alternately one could create a relation:associatedStreet across the whole settlement to reduce the amount of repetitive on the address nodes. Though it still would need to be discussed what keys should be added to the relation.

2

Yes, streets in such small settlements might indeed not have official names.

We’re using addr:housenumber and addr:place if street does not have a name (or of course addr:street instead of addr:place if it does have names). That approach was agreed on talk-hr mailing list.

Here is common example of such tagging: Way: 1103651325 | OpenStreetMap

We IMHO better should not needlessly adding addr:city=*, addr:country=* and other tags which can be derived from administrative boundary polygons (i.e. if some building is inside Relation: ‪Grad Sisak‬ (‪8778895‬) | OpenStreetMap there is no need to add database bloat addr:city=Sisak and addr:country=Croatia, as those are implied and be auto calculated), except perhaps in very rare corner cases (e.g. when split between boundaries)

About addr:postcode=*, it is IMHO hard, as I’m not sure that official database if available for use in OSM. E.g. even if some postal address might be delivered if specified 10000 Zagreb, its actual addr:postcode might be 10104 or something else instead; so lots of them are mistagged (E.g. Node: ‪Dječji vrtić Grigora Viteza, područni objekt Rujanska 21‬ (‪3366814967‬) | OpenStreetMap)
So I would avoid adding addr:postcode unless absolutely certain. And for smaller places, where there is only only one postcode, it should not be tagged on every building (due to same spam reasons as addr:country), and is better to tag postal_code on administrative boundary polygon instead (and again, only in cases when it is perfectly clear which one is applicable where, and is not an import / mechanical edit)

Well, this one should be discussed… I’m not sure it would bring much benefits, but on the other hand using name is not really correct either… Who do the other think?

In any case @barefootstache, I would suggest prioritizing other tasks then addresses in Croatia, as we’re planning on running import from official DGU sources - probably to be discussed on SotM Croatia in May 2025 (@Janjko , any ideas?)

Also, any import or mechanical edit of course would need to be discussed separately; those guidelines above are for manual mapping…