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

Martin Atkins mart at degeneration.co.uk
Sun Apr 7 20:35:23 UTC 2013


On 04/07/2013 11:58 AM, LM_1 wrote:
> In my view the streets should be more detailed - after all having the
> details dropped by computers is possible (even if not always easy). but
> detail that is not there cannot be add in any simple way. This case
> might depending on the precise conditions fulfil the requirements for
> separate direction lanes. If not some more detailed scheme would have to
> be used - mapping streets as areas. Eventually that will be the only
> viable option city centres currently mapped in high detail with only the
> streets being overly simplified...
>

I wonder if the root problem is that we've conflated the idea of the 
physical construct of a street with its parallel in the routing network.

The most complete mapping scheme would use areas to describe the 
physical area occupied by the sidewalks, street areas, boarding islands 
and other street features, and then represent the routing network as a 
separate schematic of ways on top of it with little or no visible impact 
on normal rendering. That would be very time-consuming to maintain, of 
course, and would essentially turn OpenStreetMap into a huge, 
collaboratively edited aerial "photograph" with a routing database 
alongside it. :)

It seems like the current OSM data model is really designed for and best 
to suited the low-detail schematic mapping rather than high-detail 
mapping; abutting features just manifest as ways that happen to share 
nodes, or worse: ways that happen to just sit alongside one another and 
have to be maintained individually by mappers.

An interesting thought experiment is what OSM might look like if it had 
started with a different spatial data model. For example, what if it 
were a graph of connected 2D polygons, like the map format of the Doom 
or Duke Nukem 3D game engine, or even subdividing 3D space with planes 
like the Quake game engine? That sort of model would favor realistic 
physical mapping over schematic mapping.

I wonder if it's really feasible for the use-case of highly-detailed 
renderings and the use-case of highly-accessible collaborative editing 
of a basic highway network to coexist in the same system; the former is 
something that requires extensive effort of a single person or 
coordinated group, while the latter is more suitable when you have many 
uncoordinated people who each have comparatively little time to spend.

(If Google's self-driving cars ever take off in the mainstream I guess 
we'll *all* have the hardware necessary to create an accurate 3D model 
of the world and we'll just have to figure out how to store it!)





More information about the Tagging mailing list