[Tagging] Bridges and layers

David Earl david at frankieandshadow.com
Tue Jul 27 11:27:25 BST 2010


On 27/07/2010 10:21, John Smith wrote:
> Why do taggers have to compensate for poorly written programs making
> use of the data?

Why does the data model have to make it so difficult for data consumers 
in the first place?

You cannot tell from our data model whether a bridge supports two ways 
or whether there are two parallel bridges, unless you, the tagger, says 
so (in a relation, on which there is no common agreement). e.g.: 
http://osm.org/go/0EQSjpjFr-- where the two roads are actually on 
*separate* bridges,. but the cycleway to the north shares the northern 
bridge with the road. Even if you did this as a separate polygon (which 
there is no support for in any of our software and would be a big burden 
on data producers, and doesn't take account of the hundreds of thousands 
of bridges already out there), you'd still need the relation for 
applications other than just rendering.

You *can* tell which ways go underneath a bridge without tagging help. 
But you *can't* do it fast as it requires a search and some moderately 
complicated geometry. No doubt you want fast rendering as well as 
efficient tagging.

You cannot tell from our model, without additional information such as a 
relation, whether two parallel ways are part of a dual carriageway or 
just parallel roads.

You cannot reliably tell which street a property fronts onto.

You can't even easily tell what a street is. Yes you can deduce it, but 
I now regret the loss of segments that cause streets to be broken up 
into dozens of ways. It's not the renderer that's poorly written here, 
nor the tagger, it's the data model that forces vast amounts of 
heuristic code onto the consumer just to tell what a street is.

David




More information about the Tagging mailing list