[OSM-talk] Property Address info - was Featured images for christmas week

[Andy Robinson (blackadder)]

Karl Newman [mailto:siliconfiend at gmail.com] wrote
>Sent: 22 December 2007 6:02 AM
>To: Andy Robinson
>Cc: David James; talk at openstreetmap.org
>Subject: Re: [OSM-talk] Featured images for christmas week
>
>On Dec 21, 2007 6:43 PM, Andy Robinson <blackadderajr at googlemail.com>
>wrote:
>>
>> We thrashed out ideas on th wiki the other day and decided upon the
>following:
>>
>> The buiding areas are tagged with building=residential (or
>> building=whatever if its something else)
>>
>> The nodes are also tagged with buidling=residential and have an
>> additional ref=# on them for the house number.
>>
>> Will document this in Map Features if it looks to work out ok once
>rendered.
>>
>> Cheers
>>
>> Andy
>> --
>> Andy Robinson
>>
>
>Okay, that will work for numbering buildings and showing the number on
>a map. The argument I'm having on the Wiki is more for house numbering
>which can be used for address interpolation along a street. GPS
>devices use this sort of numbering (which is my intended application).
>

Yes, its something I've noted that the house numbers are often too close to
more than one street to be sure which street they belong to on a proximity
basis. If the nodes are always within a building area then that area in the
majority of cases can be linked to its street through proximity. There are
however some exceptions, especially buildings which bound more than one
road. So it would be necessary to add specific information to either the
building or the node to identify the street in these cases but it should be
possible to get to the majority of locations without adding a pile of extra
tags. Beyond that it should be possible to use a proximity basis for all
other data (town, region etc).

There is also the issue of whether every property number is required. I find
that when using off-the-shelf navigation products they can generally take me
to the end of the correct drive. But I find I'm driving so slowly by that
point I could have worked it out anyway. So do we really need every unique
address? Probably not. For the vast majority of cases an interpolation
between the start and end number on a continuous run of property (whether
odd, even or mixed) should be enough to work out the intermediate properties
with sufficient accuracy for routing. Oddball properties or properties with
names alone will be in the database individually anyway.

Of course if the unique postal_code is added to a node or area that gives an
extra method for location resolution potential.

Cheers

Andy