[OSM-talk] Frustration of an experienced mapper

[Christoph Eckert]

Hi all,


as someone who has 700.000 trackpoints on his hard drive and who has 
mapped 70% of his hometown, Karlsruhe, Germany, I just have created a 
Garmin map of it. I use QLandkarte to view it and to detect errors 
before uploading it to the GPS device. I noticed the following issues 
which all drove me nuts each time I created a map. I spent various 
hours of fixing issues the software could have done for me. This is not 
a complaint at all, but I'd be glad if the issues below lead to an 
improved data model one day.

Segment IDs

So here we go by a (simplyfied and theoretical) first example. Given the 
fact you have four nodes A, B, C and D. The nodes are connected by 
segments of the same direction, but not in the right order, due to 
editing after their first creation. The segments therefore have the IDs 
10, 30, 20 instead of 10, 20, 30. Now I make them a way. Guess what 
weird things will happen when the street gets rendered or converted to 
a Garmin map.
I hate to say that, but I find it ridiculous that I have to manually 
edit such things to fix the database. I told my editor that there are 
four nodes and that those nodes form a street. Why the heck does the 
database assume that the street is formed of the segments instead of 
the nodes, and why does it order them by their IDs? My street consist 
of a polygon formed by its vertices, not by a certain direction or by a 
(potentially non-)consecutive number of segment IDs.

Polylines for streets and areas

Polylines would allow for curves defined by three nodes, saving a lot of 
unnecessary nodes, segments and mouseclicks to form a bow. It would 
therefore lead to a much reduced data size and allow for much better 
render results.

Oneway tagging

A similar issue like the first one is valid for oneways. Imagine a 
street formed by three nodes, A, B and C. Segments have been drawn from 
A to B to C. This forms one way. Now imagine there is a oneway from B 
to A and a oneway from B to C. Why is it not possible to tell the 
database this fact on base of the nodes? Currently I have to split the 
street into two parts. As a street usually consists of more than 2 
segments, it often is necessary to delete an existing way and some of 
its segments, to redraw the segments for oneway's sake, to make them 
ways again and to apply all the tags someone else had assigned before. 
What drives me nuts is the fact that I usually tag my stuff according 
to the issues mentioned above. But as soon as someone else tries to add 
further information, it is likely that he (incidentally) drops some of 
the data I entered before.

Currently OSM lives from collecting new data, but editing data will 
become more and more important. Thus I can imagine that we need some 
improvements on the data model, which not only address the issues 
mentioned above but even others. Someone savvy here who can contribute 
some technical/mathematical knowledge about those issues?

I hope that all made sense and the usual HTH :) ,


ce