[OSM-newbies] highway=traffic_signals

[Richard Welty]

On 4/11/10 11:55 AM, Richard Weait wrote:
> On Sat, Apr 10, 2010 at 3:14 PM, Dave F.<davefox at madasafish.com>  wrote:
>    
>> Right, OK, let me clarify my understanding of traffic lights. The nodes
>> I place represent not only the actual lights but also where the vehicles
>> are meant to halt. I see it as a type of barrier.
>> That pole is a combination of two lights for traffic travelling
>> different directions who's halts are positioned at the nodes.
>> Similar to the one in the middle of the pedestrian crossing.
>>      
> Dave, I suspect that you and I will continue to disagree on the matter
> of highway=traffic_signal.
>    
i agree with Richard. I've been out of this conversation because I've 
been traveling, and
it's permitted me to think this through.

there are two different things that can be modelled here:

1) physical location of traffic lights.

2) the location/lanes of traffic controlled by the lights.

modeling physical location conforms to a very literal definition of mapping.
modeling the traffic control conforms to a more logical definition. a 
sophisticated
routing algorithm could use logical modeling to refine a route. it would 
find
physical mapping useless as there is not enough information. the only way to
make physical mapping useful to a routing algorithm is to add enough
information to allow the logical control to be inferred, and by the time 
you've
done that, what you've really done is modeled both.

so you need to think about why you're modeling traffic control devices in
order to make the choice. is it about really nice physical rendering, or are
there perhaps other customers for the data besides renderers, say, routing
algorithms to name the most obvious one?

richard