[Tagging] RFC 2 - addr:interval

Sarah Hoffmann lonvia at denofr.de
Fri Jan 8 15:43:00 UTC 2021

On Fri, Jan 08, 2021 at 02:43:54PM +0000, Paul Allen wrote:
> There's a distinct chicken-and-egg aspect to this. Geocoders aren't going
> > to trouble to implement it until the data are there, and mappers who want
> > their addresses to be interpreted correctly by geocoders will resort to
> > enumerations with semicolons.
> >
> I think the opinion of geocoders regarding feasibility of our suggestions
> should
> be given a lot of weight.  If they say they refuse to handle something we
> should
> regard it as not being viable.

In case it wasn't clear: I am speaking for Nominatim. To summarize my
view on this disucssion...

It is very unlikely that Nomintim will ever support extended syntax in
the addr:housenumber tag. The only viable reason to have intervals in
the addr:housenumber tag is that it can be used "as is" by a data
user that wants to leave the interpretation to humans. The main data
users doing that are renderers. Any clever syntax involving brackets,
'|' signs or the like will break the addr:housenumber tag for them.

You want anything beyond a simple <from>-<to>, then a set of new tags
is the way to go. Looking at the examples provided by Kevin and Martin,
it also better suits the situations where it is not clear if something
like 1-3 is to be used as a single housenumbers or separate ones. Just


And the geocoder will happily find the address with '1-3', '1', '2' and

Even with an extra tag, parsing of complexer examples is unlikely to
happen unless there is a) a significant use with cases that cannot
be mapped as addr:interpolation or separate address points and b)
the parsing doesn't involve guessing again. The current proposals
don't really meet b) yet. (Bonus points, if it also covers alphabetic
ranges: 13a-f) And I have no idea how many cases of a) really exist.

Regarding the current interpretation of addr:housenumber, I am firmly
in the camp that its current definition of the tag is that of a single
housenumber and that those who already have been putting intervals into
it have been abusing it. So, while I'd be fine implementing the current
proposal of 'addr:interval' together with the limited version of
'<from>-<to>', Nominatim would then always interpret the absence of
addr:interval as a 'no' because that is the only interpretation that
is backward compatible with the current running definition.

(Nominatim maintainer)

More information about the Tagging mailing list