[OSM-talk] OSM the mediocre alternative

[TM]

First of all, I did at some time (sort of, see [1]) suggest that people
(actually, I think there was just one) wanting to map individual houses
could do so by using nodes. Not because it would be the right way to do
it, but because it could be done in the current model and extracted
later, when a more appropriate method had been decided upon and implemented.

At that time, there was talk of "address ranged road network files",
which seemed to be the right way to do it. That would probably require a
more static dataset, though. And, as our data is anything but static,
it's probably not going to be a solution for quite some time.

However, if a way (whether it's a list of segment or a list of points)
does not extend past junctions, then this information (basically just
four ints: low/high odd and low/high even numbers) could be added to
that way. There has been some discussion [2] about limiting ways to
topological "segments". (The current practice seems to be to avoid
branches, but not necessarily split ways because another way branches
off them.) Doing that would give you the location of an address as being
on a particular road/street between two junctions, which should be good
enough for any purpose.

    /   TM

[1] http://lists.openstreetmap.org/pipermail/talk/2006-September/007134.html
[2] http://lists.openstreetmap.org/pipermail/talk/2007-March/012309.html


Robert (Jamie) Munro wrote:
> SteveC wrote:
> > On Sun, 2007-04-22 at 12:13 +0200, Martijn van Oosterhout wrote:
> >> On 4/22/07, Christopher Schmidt <crschmidt at metacarta.com> wrote:
> >>> In OSM, that would be 88. 88 queries to the API, just so I can display
> >>> two features.
> >> That's just bad design. There's no reason why the API can't just take
> >> a list of ways and return all information (segments and nodes) needed
> >> to draw that way. The way it is setup now makes it really hard to do
> >> anything efficiently.
> >>
> >> Personally I'm for dropping segments and having ways be a list of
> >> points, sort of like a polyline, but much easier on editors...
> > This was discussed at Essen and agreed as a Good Thing. I was surprised
> > there wasn't a ticket so I added one
>
> I think that tagging segments is probably the only sensible way to
> introduce house numbers to the database. If we add a tags to the segments:
>
> numbers_left=1-7 odd
> numbers_right=2-8 even
>
> we could interpolate the other numbers in the road. If we have to tag
> the whole way with house numbers, the interpolation would be nothing
> like as accurate. Having a node for every house seems way over the top
> in most cases. By the way, I suggest that instead of a range, a list
> would be possible where a segment has non-contiguous numbering e.g.:
>
> numbers_left=1;1a;3;5;9
>
> I suppose we could instead put house numbers on nodes, but where a node
> is a junction between more than one road, this would get very awkward.
>
> A node for every house just seems way over the top.
>
> Robert (Jamie) Munro

_______________________________________________
talk mailing list
talk at openstreetmap.org
http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk