Re: [talk-au] Traffic Signals

2010-03-11 Thread Roy Wallace
On Fri, Mar 12, 2010 at 8:52 AM, Steve Bennett  wrote:
>
> ... So far, I haven't seen
> much evidence that we have ways of aggregating excess information into
> more manageable chunks.

As others have already suggested: we need relations.

There's already proposals semi-underway here:
http://wiki.openstreetmap.org/wiki/Relations

They need more work, though, so... go for it!

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread Steve Bennett
On Fri, Mar 12, 2010 at 10:21 AM, John Henderson  wrote:
> Other than de-cluttering (which tends to be done automatically anyway)
> I'm not sure why you'd want to render only one set of lights if there
> were more than that.

Well, because to most people a "set of lights" covers a whole
intersection. If there are lights northbound, southbound, eastbound
and westbound, that would be one "set of lights" to most people. You
could equally ask, why would you want to render 4 sets of lights when
there is only one?

> In this case, placing the lights accurately in their lane gives the
> correct count whereas it's your system which doubles up the number!

Heh, you could be right. I think the relation scheme David referred to
would be the way to go.

> Whether on not an intersection has lights for a particular vehicle often
> depends on the exact roure taken through that intersection.  To
> oversimplify can often be to mislead.

Often? Apart from left-turning sliplanes, are there other cases?

Steve

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread John Henderson
On 12/03/10 09:52, Steve Bennett wrote:

> At the cost of managing that extra information. So far, I haven't seen
> much evidence that we have ways of aggregating excess information into
> more manageable chunks.
>
> Put it this way: how would you render a single circle for any
> intersection that has a traffic light? That is, if there are traffic
> light nodes at one intersection, you still only want to render one
> circle. It's a pretty obvious use case.

Other than de-cluttering (which tends to be done automatically anyway) 
I'm not sure why you'd want to render only one set of lights if there 
were more than that.  But you could always limit the number to one for 
any given radius of a crossing or close group of crossings.

> How would you count the number of traffic lights along a given route?
> The scheme that's been described here would return double the actual
> number.

Consider divided roads crossing, where you'd presumably put 4 sets of 
lights on your scheme (at least that's how I inevitably see it being 
done in OSM).

In this case, placing the lights accurately in their lane gives the 
correct count whereas it's your system which doubles up the number!

> I'm not saying extra information isn't sometimes a good thing, but the
> task of simplifying that down to a useful set of information isn't
> trivial.

Why do you think anyone needs to?

> Perhaps we want to distinguish between an "intersection with lights"
> node and an "actual traffic light" node.

Whether on not an intersection has lights for a particular vehicle often 
depends on the exact roure taken through that intersection.  To 
oversimplify can often be to mislead.

John H


___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread Steve Bennett
On Fri, Mar 12, 2010 at 9:43 AM, David Murn  wrote:
>From a routing perspective, its more useful to have the information on
> the road intersection.  While it might render nicer if you put objects
> geographically where they are (separated from the road), from a routing
> perspective thats not much use unless you use a relation to relate the
> traffic lights to the roadway the traffic lights are on.

This is the general tension between a schematic diagram and an
accurate representation of the world, which has become more accute
since the arrival of Nearmap, where sub-metre accuracy now matters. It
also causes the problem with footpaths meeting roads: part of the
footpath represents the actual physical location of the footpath, and
part is a schematic line showing that it does in fact connect with the
road. But no distinction is made. It shows up with bus stops too: the
bus stop is not physically part of the road, but schematically is.

The only solution I can see is that we end up with some tags that are
explicitly schematic rather than geographically accurate, and with
some kind of relation between the physical location node and the
schematic node
Steve

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread David Murn
On Fri, 2010-03-12 at 09:52 +1100, Steve Bennett wrote:

> Put it this way: how would you render a single circle for any
> intersection that has a traffic light? That is, if there are traffic
> light nodes at one intersection, you still only want to render one
> circle. It's a pretty obvious use case.
> 
> How would you count the number of traffic lights along a given route?
> The scheme that's been described here would return double the actual
> number.

One thought that has been proposed here before, is adding all traffic
lights at an intersection to a relation.  That way, you  simply count
the number of traffic light relation groups along the way rather than
the number of nodes.  This means if you have two junctions close to each
other in distance, but that are physically separate traffic light
controls, that youd add each light of each group to 2 different
relations.


> Perhaps we want to distinguish between an "intersection with lights"
> node and an "actual traffic light" node.

The problem here, is that if you have an 'intersection with lights' that
doesnt necessarily affect all ways going through the intersection.  For
example, some roads have traffic lights  on the slip-lane where others
dont, and as I mentioned in my first email, you also have to allow for
the crossing=traffic_signals node used between foot/highway
intersections.

David


___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread John Henderson
On 12/03/10 09:43, David Murn wrote:
> From a routing perspective, its more useful to have the information on
> the road intersection.  While it might render nicer if you put objects
> geographically where they are (separated from the road), from a routing
> perspective thats not much use unless you use a relation to relate the
> traffic lights to the roadway the traffic lights are on.

A good point.  But I didn't even think of putting the lights anywhere 
other than as a node of the appropriate road.

John H

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread Steve Bennett
On Fri, Mar 12, 2010 at 9:09 AM, John Henderson  wrote:
> The benefit is in greater accuracy and completeness.  If we can do
> better than commercial street directories, then why not?

At the cost of managing that extra information. So far, I haven't seen
much evidence that we have ways of aggregating excess information into
more manageable chunks.

> Insisting that everyone do it that way would be a pain.  But if the
> information is available then I see no problem in my capturing it if
> convenient.  Nor to I see any pain caused to the user of the more
> complete maps.

Put it this way: how would you render a single circle for any
intersection that has a traffic light? That is, if there are traffic
light nodes at one intersection, you still only want to render one
circle. It's a pretty obvious use case.

How would you count the number of traffic lights along a given route?
The scheme that's been described here would return double the actual
number.

I'm not saying extra information isn't sometimes a good thing, but the
task of simplifying that down to a useful set of information isn't
trivial.

Perhaps we want to distinguish between an "intersection with lights"
node and an "actual traffic light" node.

Steve

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread David Murn
On Thu, 2010-03-11 at 23:50 +1100, Franc Carter wrote:
> Hi,
> 
> I use to tag traffic signals at the intersect of the roads, however
> with NearMap I can see that for complex intersections this does not
> work as well as I would like, three things I can see to do are:-
> 
> 1. tag at the intersecttion of roads
> 2. tag at the location of the signals
> 3. either (1) or (2) depending on the complexity of the intersection.
> 
> What's peoples views on this ?

>From a routing perspective, its more useful to have the information on
the road intersection.  While it might render nicer if you put objects
geographically where they are (separated from the road), from a routing
perspective thats not much use unless you use a relation to relate the
traffic lights to the roadway the traffic lights are on.

Also, traffic signals arent only used for road intersections, traffic
lights are also sometimes used to control pedestrian crossings between a
footway/cycleway and a highway, which I feel is useful to tag.

David


___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread John Henderson
On 12/03/10 08:39, Steve Bennett wrote:
> On Thu, Mar 11, 2010 at 11:55 PM, Luke Woolley  wrote:
>> Personally, since NearMap became available, i've been placing traffic light 
>> tags at every thick white stop line at the intersection, which means for a 
>> standard intersection, there are 4 traffic light nodes.
>
> What's the benefit of this rather than, say, one traffic light node on
> the centre of the intersection? Is the exact location of the light
> itself important? I'm comparing this approach to the Melway, which
> uses a big purple circle across the intersection to indicate the
> presence of traffic lights, mostly as a landmark I guess.

The benefit is in greater accuracy and completeness.  If we can do 
better than commercial street directories, then why not?

To my mind, it's a bit like deciding whether to make a minor road with a 
concrete dividing strip a divided road or just a two-lane, two-way 
single road.

> The extra detail of individual lights might be more of a pain in the
> arse than actually useful?

Insisting that everyone do it that way would be a pain.  But if the 
information is available then I see no problem in my capturing it if 
convenient.  Nor to I see any pain caused to the user of the more 
complete maps.

And you never know when accurate information might prove useful.

John H

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread Steve Bennett
On Thu, Mar 11, 2010 at 11:55 PM, Luke Woolley  wrote:
> Personally, since NearMap became available, i've been placing traffic light 
> tags at every thick white stop line at the intersection, which means for a 
> standard intersection, there are 4 traffic light nodes.

What's the benefit of this rather than, say, one traffic light node on
the centre of the intersection? Is the exact location of the light
itself important? I'm comparing this approach to the Melway, which
uses a big purple circle across the intersection to indicate the
presence of traffic lights, mostly as a landmark I guess.

The extra detail of individual lights might be more of a pain in the
arse than actually useful?

Steve

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread John Henderson
On 11/03/10 23:55, Luke Woolley wrote:
> Personally, since NearMap became available, i've been placing traffic light 
> tags at every thick white stop line at the intersection, which means for a 
> standard intersection, there are 4 traffic light nodes.

This is exactly what I've been doing too, although the intersections 
I've done this way are all ones I'm familiar with.

John H

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


Re: [talk-au] Traffic Signals

2010-03-11 Thread Luke Woolley
Personally, since NearMap became available, i've been placing traffic light 
tags at every thick white stop line at the intersection, which means for a 
standard intersection, there are 4 traffic light nodes.

On 11/03/2010, at 11:50 PM, Franc Carter wrote:

> Hi,
> 
> I use to tag traffic signals at the intersect of the roads, however
> with NearMap I can see that for complex intersections this does not
> work as well as I would like, three things I can see to do are:-
> 
> 1. tag at the intersecttion of roads
> 2. tag at the location of the signals
> 3. either (1) or (2) depending on the complexity of the intersection.
> 
> What's peoples views on this ?
> 
> thanks
> 
> -- 
> Franc
> 
> ___
> Talk-au mailing list
> Talk-au@openstreetmap.org
> http://lists.openstreetmap.org/listinfo/talk-au


___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au


[talk-au] Traffic Signals

2010-03-11 Thread Franc Carter
Hi,

I use to tag traffic signals at the intersect of the roads, however
with NearMap I can see that for complex intersections this does not
work as well as I would like, three things I can see to do are:-

1. tag at the intersecttion of roads
2. tag at the location of the signals
3. either (1) or (2) depending on the complexity of the intersection.

What's peoples views on this ?

thanks

-- 
Franc

___
Talk-au mailing list
Talk-au@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk-au