[OSM-talk] House numbers... One more suggestion
siliconfiend at gmail.com
Mon Jul 28 19:37:04 BST 2008
On Mon, Jul 28, 2008 at 11:15 AM, Frederik Ramm <frederik at remote.org> wrote:
> Karl Newman wrote:
>> Given a street, I'd have to look through NN million nodes to find the
>> closest ones
> No. The nodes are tagged with street names, thus duplicating info found on
> the way. You just need an index on that field, look up the street name in
> the index, and immediately know all addresses entered for that street.
Only for some. Several examples show only the address, not the street. Plus,
even if it's used, the street association is still subject to typos, data
corruption, duplication, etc. Tying the node to the street by using its name
is really poor for normalization.
> THEN, if you want to know how many meters from house number 13 to house
> number 723, you have to compute the nearest point on the nearest way for 13
> and 723 and find the distance.
> That being said, some people actually add relations to make things extra
> clear (see bottom of KS page). The reason we don't force people to add
> relations is that it seemed too much to ask - we'd rather make it a bit more
> difficult for the data user than for the mapper. The Karlsruhe Schema was
> made with mappers in mind, who enter house numbers manually. Had it been a
> schema designed for importing data from another source, we might have opted
> for a more complex data model, saying that nobody has to enter it anyway...
It's not "a bit more difficult", it's a huge burden. It means you presume
the data consumer has to be working from a database that contains the entire
area you're interested in (maybe the planet?). Your suggestion demands
random access to the entire dataset and thus precludes using pipeline tools
such as Osmosis.
How is it easier on the mappers if they have to go back and enter the street
numbers twice if they want to be able to use it for something other than
drawing numbers on a map?
> But as I've said a number of times - I wouldn't feel the least offended if
> someone else tried another scheme.
Many people have made numerous alternatives (8 (!!!) suggestions on the
proposal page, including mine), and they can't seem to see the light... all
of the numbering schemes I've seen (TIGER, Garmin, local city & county GIS,
etc.) have the address numbers associated with a node in a way and a
left/right/both scheme. It's just unfortunate that people are being referred
to the inadequate Karlsruhe schema as the de facto method for entering
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the talk