[mkgmap-dev] RFC: naming unnamed roads

Sun May 3 18:36:28 BST 2015

On Thu, Apr 30, 2015 at 09:57:04AM -0400, Greg Troxel wrote:
>>The general assumption would seem to be that the street names attached 
>>to house addresses belong to roads that are reachable by car, or that 
>>each residence is reachable by car. Maybe in some rare case there is 
>>some access restriction on the road associated with the address, such 
>>as access=destination. There could be named cycleways or footways 
>>between the road and the address node, but no named public roads with 
>>a different name, unless there is an error in the map data.
>
>That's an interesting point.  In the US, around me, there really aren't 
>such assumptions.  Instead, a lot (area of land that can be bought and 
>sold as a unit) has an address, generally taken from a public or 
>private way that borders the lot.

I see. I have some knowledge of the administrative system of properties 
in Finland. I do not think that street names play any role there. A 
property may have been assigned an arbitrary name by its owner, but it 
always has a registration number that uniquely identifies the lot. If 
the lot is split, I guess that all parts of it will get a longer 
registration number (adding some suffix to the original number).  
Nowadays, the fully qualified registration number should start with the 
NUTS (Nomenclature of Territorial Units for Statistics) prefix of the 
municipality.

Streets may be renamed or renumbered (seldom, but possible when lots of 
properties have been split and new houses built, for example). This 
would not affect the registration numbers of the properties. BTW, I 
don't think that the registration numbers will or should be put to the 
OSM database. They are only relevant when buying or selling properties.  
It should suffice to have the administrative boundaries of the suburbs 
or villages.

In Finland, when the same street goes through multiple municipalities, 
it usually changes names at the border. This is unlike what I saw in the 
US: El Camino Real in California would be called that for the whole way, 
and you would have to pay attention when crossing borders, because there 
could be multiple identical house numbers, maybe some tens of miles 
apart. Sure, in every country, you will have common street names such as 
"Main Street" or equivalent in every town, but those would typically not 
be part of a major route. However, this should not be a problem for the 
Garmin format. El Camino Real would be split to sections at municipality 
borders, and each section would be assigned its own house number data.

>That's a long way of saying that it's messy and that general rules 
>don't always hold (in mass.us; not saying that applies to .fi).

Right. We need to identify the least common denominator, and also need 
to find ways to circumvent some limitations of the Garmin format. I 
think that the simpler, the better, as long as we are not introducing 
too many artifacts, such as moving a housenumber to a street far away.

>This is quite separate from access.  There are addresses on military 
>cases, and very often in residential complexes with gates that you need 
>a code/etc. to get through.   So it's not just access=destination but 
>access=private, and yet they are real addresses on named roads inside 
>the gate.

Oh, yes. I would not worry too much about access. From routing point of 
view, there should be not much difference between access=private and 
access=destination. The routing engine can tell you how to get to a 
location. It cannot know if you are really allowed to go there. All that 
matters is that it should not suggest a restricted road as a shortcut 
for some other location that is reachable via other roads.

	Marko