[OSM-talk] Suggestion for JOSM

Nop ekkehart at gmx.de
Mon Nov 2 07:45:01 GMT 2009


Hi!

Shalabh schrieb:
> Given my limited understanding of mapping and even more limited 
> understanding of computing, I think it would be better if JOSM assumed 
> the trails to be correct and drew nodes on it on its own.

This is possible combining several features. However, it is a bad idea, 
most ways added this way are in horrible state and need correction.

- way too many nodes. The API does not return more than 50000 nodes in 
one request, so many tracks with 1500 nodes each quickly make it 
impossible to download a sizable area
- the GPs is inaccurate. If you just stupidly add the track, all the 
mismeasurements are added to the DB. While drawing the way, you can 
smooth out the obvious zigzags of errors and deviations of known bad 
reception. Also, as the distance of nodes (usually 1m) is way smaller 
than the basic error of the GPS, it makes no sense to add this sort of 
misleading pseudo-accuracy
- those ways are then unconnected to all other ways. It is very 
difficult and tedious to create the proper connections
- most people who take this "easy way" don't connect and simplify the 
way properly, you will often find ways that are simply created over 
existing, manually edited versions of the same way.

So in practice this doesn't work out. If you process your track 
properly, it is quite some work either way, but using the track directly 
encourages quick and sloppy adding of bad geometry.

It has been suggested several times, that the possibility to do this 
indirectly be removed from JOSM altogether and having corrected many bad 
direct uploads I am rather in favour of this.

bye
	Nop




More information about the talk mailing list