[OSM-talk] Finding a particular street on the GPS

[Karl Newman]

On Dec 24, 2007 2:49 AM, Christoph Eckert <ce at christeck.de> wrote:
> Hi,
>
> > I don't think so--I think it's the full, unrestricted version with
> > routing and everything.
>
> http://mapcenter.cgpsmapper.com/faq.php
>
> Q: What are the advantages of the online compiler? Is it better than
> freeware/shareware cGPSMapper?
>
> A: In addition to all features of freeware version, online compiler supports
> all features available in registered shareware version (city and POI
> indexing, additional city information: country and region, additional POI
> information: country, region, city and description), some features of the
> standard version (direct support for ESRI format, irregular map bounds, not
> limited city and POI indexing, full POI address and additional description),
> some features of the professional edition (building numbering, searching for
> address, searching for intersection) and even routing
> (with some limitations).
> ^^^^^^^^^^^^^
>
> I'd like to know what "some" contains.
>

Ah, I hadn't seen that. The restrictions most likely have to do with
the size of maps you can create. Well, anyway, it's good for testing.
The long-term solution would be to get a license for the routable
version and be done with it.

> [...]
>
> > Yes, absolutely, but everyone keeps trying to make a single relation
> > into a number range, instead of the simpler and more flexible case of
> > numbering single points with one relation at each point where you want
> > to provide a number, with the number range between the points implied
> > based on the numbering scheme indicated
>
> I think the range idea is the usual way osm works: first map roughly, then
> refine it as time passes by. IMO we will not only see a node per number, but
> even a polygon for each building with a number applied. I still think a node
> per number makes a lot of sense, even if the node was part of the
> aforementioned polygon. This way one could tag the entrance of a house as the
> number. It will help to create routing exactly to the entrance, even if a
> house has more than one number/entrance.

No the range idea will lead to duplication of data (at the range
endpoints--the number will appear in two different relations). It's
more prone to errors and will lead to complications when the ranges
start to overlap. If you were a new mapper and wanted to add street
numbers, you would first have to go find all the relations and sort
out the street number ones to figure out where to add yours. Too
complicated, and unnecessary. With my proposal, it's not necessary to
number every address (interpolation will fill in the blanks), but it
is possible.

Keep in mind what we're trying to accomplish here--we're not trying to
drop a bomb on a street address, we're trying to locate the relative
position of that address along its street (indicating which side of
the street it's on is a bonus, too).

>
> > (read my proposal for more
> > details--it's really quite simple).
>
> Sorry for my lazyness, but can you provide an URL? Didn't find it in the wiki.

Sorry. I posted it earlier in this thread I think, so I thought you
had it. Anyway, here it is:
http://wiki.openstreetmap.org/index.php/Proposed_features/House_numbers#Third_Suggestion_.28Using_relations.29

>
> > This is the format similar to what
> > cGPSMapper requires, and it makes a lot of sense. The GPS
> > manufacturers solved this problem already; let's not re-invent the
> > wheel.
>
> IMO it's somewhat Chicken-or-the-egg. If routers used it, people would map it
> like recommended, and if people mapped it like recommended, routers could use
> it.

It's not really chicken-and-egg. The egg's already laid; the chicken
is hatched. If we hope to use this on existing navigational devices
(i.e., Garmin GPSr as this thread was about), then it needs to be able
to easily translate into the format it requires.

Karl