[OSM-talk] Some question about OSM

[Simon Hewison]

Nicola Ranaldo wrote:
> I'm writing a qt4-based OSM client in order to render maps and doing route 
> planning with speech instructions (festival). It has some basic editing 
> capabilities too.  Depending on my free time, when the code will be cleaned 
> and optimized, i'll release it under gpl.
> While approaching i collected some technical questions for you:
> 
> *) wiki says a segment could belong to several ways, is there a specific 
> reason for that?

Yes. A way could be an abstract concept like a bus route, or it could be a 
named street.

Think of a way as a method of adding additional information to the segments 
it includes. Route planning software should not just blindly follow a way, 
especially if it's at a junction, though a user may have a preference not to 
deviate from a signposted route descriptions (eg, they may not want to leave 
the motorway for a few kilometres along a narrow country lane, around 
complicated junctions the the like in order to cut a corner), in which case 
the number of changes of "ref" key/values along a possible route might 
weight the decision taken.

> *) osmplanet data seems to contains some dirty informations, tests, orphaned 
> objects and so on... is it normal?
> 
> *) wath's the best approach to do street and place names rendering? 
> disposition of text will be evaluated by software in real-time? will be some 
> key/value helping rendering software ? (for example, a tag could say render 
> this town names at x,y, or <important> do not render it when scale > z and so 
> on)

Render town names at x,y won't make much sense if you've got one of those 
systems that rotates the map to suit the direction of travel, else your big 
long word will obscure other things when you spin the map around.

> *) my route planning code actually works on every raw segment/node, is this 
> right? should it run only on "junction" nodes? should the software use the 
> first option and take help from the second when available? this questions due 
> to lack of junctions tagging in data , i think routeplanning helpers should 
> be strong formalized, in order to modify editors and avoid huge maps 
> reediting

For routing decisions, it needs to consider the junctions, and the links 
between junctions having a number of values, including distance, 
restrictions (one way, low bridges and the like), so you'll probably want 
your own data structure to represent them.

> *) do you think distibuting (with zone delegation ala DNS) database may be an 
> idea (with rest api modifications) ?
> 
> *) a stupid question, some segments have k="highway" v="primary", and others 
> k="class" v="primary"... is it normal?

Things are in a state of change at the moment, with some people using 
highway=primary, and other using class=primary. highway=primary is the newer 
scheme, but both should be counted as equal at present. It's all down to 
what editors people are using, and how easy it makes it to define the 
key/value pairs.

--
Simon Hewison