On Thu, 13 Aug 2009, Pascal Neis wrote:
please, could you try it again?
For me it works, http://bit.ly/HDgbO
The coordinates you've gone to there are different from the ones I gave in
my original report - you've gone to -4.1307375,53.1501034 whereas I used
-4.119322,53.145316. I have just retested and now get an error that there
isn't a street within 300 metres, which is rather more sensible than the
route I was given before.
It seems, that it was a OSM data problem, thx!
I'm not sure how this can be put down to an OSM data problem - the
destination point should have been projected onto the nearest road (i.e.
something that wasn't highway=road, since OpenRoutingService doesn't
treat these as roads). This isn't what happened - instead it sent me on
some crazy route via Europe and hundreds of miles into the middle of the
North Sea. This has now been fixed by disallowing destinations over 300
metres from a road.
The underlying problem is still there though - for example, this route
projects the destination point (just North of Swansea, Wales) onto a
shipping lane in the middle of the Irish sea instead of a nearby
highway=tertiary:
http://data.giub.uni-bonn.de/openrouteservice/index.php?start=-3.9440128,51.680206end=-3.946695,51.6843037pref=Fastestlang=denoMotorways=falsenoTollways=false
So there is something rather wrong with the algorithm that finds the
nearest road to project the destination point onto.
the highway=road tag is a a temporary tag, or?
A road of unknown classification. This is intended as a temporary
tag to mark a road until it has been properly surveyed. Once it has
been surveyed, the classification should be updated to the
appropriate value. [1]
Yes, it is a temporary tag to allow roads to be added without being
properly surveyed (e.g. I know I drove down here, I have a GPS track but
I didn't spend the time to note down what the road was like so I can't
give it a classification). It is used quite a lot by people who collect
tracks as they drive around, even when they can't spend the time to
properly survey an area. It was added to prevent unknown
classifications being added as highway=unclassified (which is what was
happenning a lot previously).
There might be some sense in trying to avoid routing people via
highway=road since we don't know what they are like (could be anything
from a tiny single track road to a big dual carriageway). However, where
the only possible way to reach a destination is via a highway=road or
highway=track, the routing algorithm probably needs to allow this.
i.e. if you are trying to get to a farm that is connected to the road
network by a dirt track (highway=track), you want the route to take you
down that track - this isn't an error in the data that should be fixed.
The only other alternatives are:
1. Tell the user that their destination is unreachable - this is wrong.
2. Project the destination onto the nearest road that isn't a
highway=track - this could lead the user onto completely the wrong road,
rather than the road that has the track connected to it.
--
- Steve
xmpp:st...@nexusuk.org sip:st...@nexusuk.org http://www.nexusuk.org/
Servatis a periculum, servatis a maleficum - Whisper, Evanescence
___
talk mailing list
talk@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk
