[Tagging] Mismatched Level of Detail in highways vs. other elements

Martin Atkins mart at degeneration.co.uk
Sun Apr 7 18:37:20 UTC 2013


Hi all,

I do mapping in San Francisco, CA and I'm frustrated about the 
inconsistent levels of detail we typically use when mapping urban 
environments.

For example, most highways are mapped in a network-oriented fashion with 
one string of ways representing both directions of traffic, often 
encapsulating other features like cycle lanes and sidewalks, and 
intersections simply represented by crossing the streets at a single 
common node.

On the other hand, rail lines are most commonly mapped by their physical 
shape, so the rail ways come in pairs. The people who mapped the tram 
lines in San Francisco also mapped the curves of the rails at 
intersections, rather than having them meet at a single node as with the 
highways. This creates the following ridiculous effect in rendering:
     http://osm.org/go/TZHvFT5aF--

Notice how the rails only just fit inside the rendered street on 
straight sections, and cut the street corner completely at the intersection.

However, here's how it actually looks on the ground (looking across the 
intersection from east to west). Notice that the rails are completely 
contained within this 4-lane intersection (all four being normal traffic 
lanes with no physical separation except for the tram boarding platforms):
     http://oi45.tinypic.com/w6qsgh.jpg

(On the plus side, we're doing better than Google Maps, whose rendering 
makes it look like the rails on Church street are both off to the west 
side of the street! http://tinyurl.com/cedot4n )

This problem shows up in various other contexts too: it's impossible to 
accurately tag a bench or bus stop on a sidewalk because the sidewalk 
doesn't exist as a separate construct. Fences or buildings directly abut 
the street end up rendering either over the street or set back from it 
because the true width of the street is not represented.

For most normal street mapping and vehicle routing purposes it seems 
sufficient to just know simple landmark details that aid in orientation, 
e.g. that whether particular street contains a railway or it passes 
alongside a railway. Of course, more detail-oriented uses like 3D 
renderings it'd be more important to have the full physical street 
layout described, with separated lanes and proper physical relationships 
with surrounding objects.

How have others resolved this fundamental conflict? More detailed 
streets, or less-detailed everything else?




More information about the Tagging mailing list