[OSM-talk] Layer transitions

[Roy Wallace]

On Fri, Aug 7, 2009 at 5:39 PM, Lambert Carsten<lhc.osm at solcon.nl> wrote:
> Adding a little piece of road so the junction can be on one layer just does
> not make sense. In Amsterdam there are lots of bridges and canals.
> The canals there are physically not on the same layer as the road and bridges.
> But for practical reasons we only add layer tags where ways cross without
> connecting (bridge over water). This 'T-junction rule' is causing just about
> every bridge to have small extra bits added. Or have roads that do not cross
> anything tagged as layer=1.

I don't understand. The T-junctions in question involve junctions of
roads and bridges, right? Aren't the roads and bridges on the same
layer? If so, they should be tagged as such (either implicitly as
layer=0 and the canals with layer < 0, or explicitly as layer=1 with
the canals as layer < 1). Right? Please give a concrete example of a
junction that's raising the question.

> The renderers already seem to have a problem with the names of roads when they
> are split up into bits. Adding these extra little bits will just make things
> worse. It seems to me that those rendering problems (at least with bridges)
> would be much better solved by looking at whatever the bridge is crossing and
> use that to decide how to render a bridge that creates a T-connection with a
> way. And if Keepright can see that a bridge or tunnel is connected to a
> T-junction why can't a renderer come to the same conclusion and render
> differently?

Forget about the renderer for a second. IMHO we should tag what is on
the ground. As the wiki says, layer is "used to mark if a
way/node/area is above or under another one", where "is above or
under" clearly refers to the physical reality. It's not *physically*
possible to have something on layer=0 connected at a T-junction to
something on layer=1 - or at least, the physical meaning of it has not
been clearly defined, as far as I'm aware (let me know if it has been
defined). Thus, in order to get rid of the T-junction-same-layer rule,
you'd have to come up with a description of the physical implication
of a T-junction of ways on different layers, and change the definition
of the layer tag on the wiki accordingly.

Alternatively, perhaps you are suggesting the layer=* tag should be
used only to tell the renderer which ways to display on top of other
ways? If so, I don't like this because it's clearly tagging for the
renderer.