[Tagging] [Talk-us] stop signs

[Anthony]

On Tue, Oct 26, 2010 at 1:36 PM, Peter Wendorff
<wendorff at uni-paderborn.de> wrote:
> Am 26.10.2010 16:50, schrieb Anthony:
>>
>> On Tue, Oct 26, 2010 at 10:45 AM, Nathan Edgars II<neroute2 at gmail.com>
>>  wrote:
>>>
>>> On Tue, Oct 26, 2010 at 10:27 AM, John Smith<deltafoxtrot256 at gmail.com>
>>>  wrote:
>>>>
>>>> On 27 October 2010 00:17, Nathan Edgars II<neroute2 at gmail.com>  wrote:
>>>>>
>>>>> What's wrong with something like highway:forward=stop or
>>>>> highway:backward=stop for the node where one must stop?
>>>>
>>>> Editors won't honour that sort of detail, so if the direction of the
>>>> way is flipped for what ever reason than the direction added for signs
>>>> becomes erroneous.
>>>
>>> I agree to some extent, but we already have other tags such as
>>> cycleway:right and relation roles of forward/backward.
>>
>> That's probably acceptable, as the errors could rather easily be
>> detected by applying the "nearest intersection" (perhaps along with
>> some sort of nolint code for those possible few intersections which
>> don't obey the standard rule).
>
> But why do you move the problem from the applications (pre)processing the
> data to editors and bug-finding tools?

I guess because assuming the stop sign faces away from the nearest
intersection is only a guess.  It may be right 99% of the time, but it
might be wrong that other 1% (for instance, there are stop signs which
are not near intersections at all).  A bug-finding tool can point out
that 1%, and then someone can determine manually whether it's a bug or
just a strange situation.

> A routing/navigation application has to calculate his routing graph from the
> data. This app can decide if it's useful to know the role of a stop sign on
> a road or not.

Routing/navigation isn't the only use for stop sign information.  In
fact, I'd say it's one of the least useful places for the information.

> Why do you want to tag it in the data, if it's not needed?

I feel it is needed, for the reasons I gave above.

> As: if it's
> possible to detect these errors automatically it should also be possible to
> calculate the "missing" data when constructing the routing graph.

Right, that's where the 99%/1% thing comes into place.  I don't think
it's true that 100% of stop signs face opposite the nearest
intersection.  So there has to be a tag for those exceptions.

Now, you could say that the tag is only needed for those exceptional
cases.  But then you still need editor support for those exceptional
cases.