[josm-dev] Validator

[Dermot McNally]

2008/8/20 David Earl <david at frankieandshadow.com>:
> On 20/08/2008 10:24, Dermot McNally wrote:
>>
>> Layers exist to
>> determine the drawing order of overlapping elements, nothing more.
>
> ... which rather violates the "don't tag for the renderer" maxim, yes?

Not at all. We gather map data to describe on-the-ground data. As our
focus is a plan-view map, we mostly don't concern ourselves with the
elevation of elements. However, since we do care about the topology of
the elements we store, we choose to care about the order in which
layers occur.

We could decide not to care about these things - a vehicle routing
solution arguably doesn't care whether a road crosses above or below a
river, but it does want to know that you can't turn from road _to_
river. (or to another road it crosses without intersecting with, just
to keep the example valid). In this regard, you can argue that, in
storing the layering sequence for overlaps, we're collecting
information that's only useful to a rendered map. But guess what?
Rendered maps are also a desired output of the project, so collecting
layering data that allows us to draw correct maps is no more
renderer-oriented than deciding to have different grades of road that
a renderer will subsequently draw in different colours or widths.

> There's a few cases where it can't be avoided, like where a bridge goes over
> another bridge in a complex intersection or by chance, as here:
> http://www.openstreetmap.org/?lat=52.329752&lon=-0.192132&zoom=18&layers=B00FTF

Or here:
http://www.openstreetmap.org/?lat=53.38212&lon=-6.36322&zoom=16&layers=B00FTF

> But for the vast majority of cases, we ought to be able to tell without
> using layers.

Perhaps - but if we do, it will be through implicit layers, which
might well be -1 for waterways and 1 for bridges.

Dermot

-- 
--------------------------------------
Iren sind menschlich