[mkgmap-dev] Access handling trunk <=> mergeroads branch

Sat Nov 30 07:32:06 GMT 2013

Am 30.11.2013 02:14, schrieb Greg Troxel:
>
> WanMil <wmgcnfg at web.de> writes:
>
>> I am comparing differences between the mergeroads branch running the
>> compatibility mode and the trunk.
>>
>> Having the following way:
>>   access = destination
>>   highway = path
>>
>> The access bits are set as:
>>                 trunk | branch
>> no_car       |   0   |    1
>> no_bus       |   0   |    1
>> no_taxi      |   0   |    1
>> no_foot      |   0   |    0
>> no_bike      |   0   |    1
>> no_truck     |   0   |    1
>> no_throughrt |   1   |    1
>> delivery     |   0   |    1
>> emergency    |   0   |    1
>>
>> 0 = bit is not set
>> 1 = bit is set
>>
>> For delivery and emergency I am not sure if delivery/emergency is
>> allowed if the bit is set or unset.
>>
>> At the moment the trunk allows all type of vehicles to use the given
>> way. I think that's not good. The branch allows only foot to use the
>> way.  What do you think should be the correct access bit mask for the
>> given way?
>
> I see multiple interlinked questions here.
>
> One is if a given vehicle can physically use the way.  For highway=path,
> I would say that car, bus, taxi, truck, delivery, emergency cannot.  If
> they could, it would be highway=track instead.  So some of these bits
> should not be set even for highway=path with access=yes (default case).
> Or perhaps I'm confused and there is some other mechanism for this.

I think the access tagging in OSM is confusing.
The wiki page tells highway=path implies motor_vehicle=no and 
emergency=destination.

For highway=footway the wiki tells that foot=designated is implied and 
several more country based access-Restrictions and maxspeeds. So to be 
accurate we have to implement 19 country specific rules just for 
highway=footway!

I think we should use reasonable values in the default style which use 
the default access mappings documented in the wiki: 
http://wiki.openstreetmap.org/wiki/OSM_tags_for_routing/Access-Restrictions#Default.
So highway=path is by default access=no and foot=yes, bicyle=yes, horse=yes.

>
> The other question is access=destination.  That's supposed to encode a
> legal notion that one may traverse the way only if it's necessary to get
> someplace that you can legally go.   My understanding of the Garmin
> routing behavior is that it will not use disallowed ways in a route,
> except at the ends.

There is a no_throughroute flag. So for access=destination the 
no_throughroute flag is set which tells Garmin not to use the way for a 
route that doesn't end on this way (as far as I understand).

>
> So what this leaves me failing to grasp (and which I think is at the
> heart of deciding the right answer to your query) is how these no_* bits
> relate to a) physical feasibility, which can not be overridden by
> 'necessary to get to destination'  and b) permission, which seems to be
> overriden.
>