[OSM-talk] Rendering (was: cadastre, ways, rendering)

[Ed Davies]

Marcus Bauer wrote:
> To avoid this you either add more and more rules to the renderer, ad
> infinitum, 

Agreed, that's pretty much impossible in the most general cases.

> or you help it a bit. 

Yes, the renderers need some help - they should get as many of the
common cases right as possible, of course, but, in the end, there
are just too many special cases for them to always work well.

> You don't loose any information if only one side of a motorway has a
> ref=A5 key/value.

But one obviously does lose information: the identity of the
untagged side of the motorway.

Expanding only slightly on what others have said: I think the aim
should be to make as accurate a model of the real world as possible
while keeping the data entry simple and robust.  Given that model,
hints can then be added to help the renderers for the corner cases
that their normal algorithms don't deal with well.  These should
be separate from the basic data model - sort of the like the idea
  of trying to keep content and styling separate in HTML and CSS.

The HTML/CSS analogy is a bit weak as here we're talking about
rendering hints mixed in with the main data but there's still a
separation principle worth keeping in mind.  Practically, I think
this means that the set of tag names for the basic data and the
set for the rendering hints should be disjoint.  You should be
able to take a .osm file and delete the rendering hint tags
without changing the fundamental meaning of the data - just
changing how nice a map would look.

Ed.