[Tagging] Housenumber interpolation with regularly skippednumbers

[Anthony]

On Sat, Oct 10, 2009 at 9:35 PM, Randy Thomson <rwtnospam-osm at yahoo.com> wrote:
> Anthony wrote:
>> If you know where
>> all the individual houses are, why bother with an interpolation?
>
> My answer to your first question is a labor saving of anywhere from 2
> to 10-fold for the mapper. I'm talking about thousands of houses.
>
How are you going about verifying that every single one of those
thousands of houses are numbered according to the scheme, that none of
them have multiple family members, etc?

>> What is one supposed to do if the number of houses is unknown?  Do I
>> need to propose an
>> addr:interpolation_as_the_word_is_defined_by_everyone_in_the_gis_world
>> _except_osm to handle that situation?
>
> My answer to your second question is you don't know anything about the
> reality of the addresses you are tagging you probably shouldn't be
> tagging them.

I didn't say I don't know anything.  I said I don't know the number of
houses.  The way interpolations work, everywhere except for OSM, is
that you record the address of the first house on the block and the
last house of the block (possibly on each side), and you assume that
the houses in between have numbers in between, roughly evenly spaced.

That's not exact, and if you have more information it's certainly
better.  But it's good enough to allow routing software to get you to
the right block, and approximately to the right section of the right
block.

And for retail areas, like individually addressed stores in a strip
mall, it may not even be feasible to find the address of each
location, even if you walked by each store (you might have to go
inside and ask in order to find the address).  And the information is
going to change more frequently than you can keep up with anyway.  Two
store locations with two addresses get combined into one.  One store
location with one address gets split into two.  This happens with
houses as well, though not as frequently.

I'm not saying it's wrong to give the extra detail, but I am saying
it's wrong for data users to assume that an interpolation from 1-11
necessarily has a 1, 3, 5, 7, 9, and 11.  Not that I can think of any
reason a data user would make that assumption in the first place.