[Talk-us] using address interpolation with building polygons

Richard Welty rwelty at averillpark.net
Tue Mar 29 19:54:14 BST 2011


On 3/29/11 1:34 AM, Val Kartchner wrote:
> On Mon, 2011-03-28 at 18:07 -0400, Richard Welty wrote:
>> ... any solution i can come up with strikes me as making things messier.
>>
>> richard
> Here's a useful question: How are addresses stored in the database for
> GPS devices?  They have to translate an address into a coordinate
> somehow.  If we answer this, then this could lead to an answer to this
> question.  Do we have anyone here that translates the OSM data to GPS
> files?
ok, here's how i see the issue:

we can attach address information to both nodes and to building polygons.

address interpolations are ways, by definition they can only have nodes.
the best we can do with building polygons is include one or more nodes in an
area interpolation, and hope that whatever consumer is processing the 
interpolation
knows to "look up" to see the address. this would be rather ugly, and i 
doubt that
any current data consumer does this.

the other alternative is an extension of the associatedStreet relation 
to serve as
a type of address interpolation container. on the one had, this would be 
pretty
easy to define. on the other hand, in a world of somewhat fragile 
relations, i think
that associatedStreet is more fragile than most.

for the case where there is an occasional building outline with a good 
address in
an area where otherwise there are address nodes with an interpolation 
way, the
explicit address override behavior that Ant described is reasonable.

this still leaves the challenge of denser areas where building outlines 
are being done,
but address info is incomplete.

richard




More information about the Talk-us mailing list