Hello,
since you are going to discuss the address import tomorrow at the
OpenGisData.eu mappers meeting, and I can't be there in person, I'll give
you some comments here:
1. "Less is more"
* addr:full tags
I'd just drop the addr:full:* tags, they are of no use here.
* addr:country tags
Drop the addr:country tags, too. This is overly redundant in my opinion.
The only place where this data may come in handy would be for addresses
that are very near to the border, but in those cases the addr:country
could be added by the mapper during the import, if one thinks that it
makes sense.
* addr:source tags
Drop those. It's good enough to tag the source on the changesets (and the
import wiki pages). No need for a source tag on every imported datum.
2. Address tagging
I think we've talked about this already [1], but addresses in South Tyrol
are of two different kinds: a) street-based or b) place-based.
If it is street-based, the addr:street tag must be used (and if available,
an addr:hamlet can be used optionally to indicate the "Fraktion/frazione"
of the address). On the other hand, if an address is place-based the
addr:place tag must be used and there must be no addr:street tag.
Looking at the test data for Martell, I see that all addresses are
converted as street-based ones, which is incorrect as most (or all?) of
them are actually place-based addresses. (See how the addresses in Martell
are currently tagged.)
I know that the WEGE-DB doesn't directly have this data (and that the data
is quite inconsistent from one municipality to the next). But that doesn't
mean that you can import this incomplete data into OSM. The solution to
this problem is either to program some kind of heuristic that takes nearby
places into account, or to let mappers decide before or during the import
(this approach would require a more elaborate importing procedure rather
than just providing 3 .osm files to import with JOSM).
3. "addr:aa2osmid"
I think we've talked about this, too. Why do you need other ids to
represent already unique identifiers? While having external IDs (like GTFS
stuff, TMC, Gemeindekennzahl, etc.) in OSM is quite bad in principle, this
particular ID is even worse: There is no accessible external dataset to
which these IDs correspond to, nobody other than you can check if they are
still correct, no mapper can know what to put in if new addresses are
mapped on the ground or if errors are being corrected, etc… Also, see
point number 1 above (less is more).
I don't see how this field should be necessary in the future. Please show
me one use-case where this ID is necessary. Until then I'm strongly
against importing this field.
4. Announce internationally
Please note that you will have to announce this import also on the
international @imports [2] mailing list. In the last few months there have
been several other address import proposals which were all heavily
discussed.
Best regards,
Martin
[1]
https://wiki.openstreetmap.org/wiki/AltoAdige_-_Südtirol/OpenGisData_HouseNumber_Import2#OSM_tags
[2] https://lists.openstreetmap.org/listinfo/imports
_______________________________________________
Talk-it-southtyrol mailing list
Talk-it-southtyrol@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-it-southtyrol
