[mkgmap-dev] Virtual paths

Tue Jul 28 22:36:23 BST 2020

On Tue, Jul 28, 2020 at 08:46:02AM -0400, Greg Troxel wrote:
> ael <witwall3 at disroot.org> writes:
> 
> > Not just mkgmap. It is a general problem. Maybe OSM should introduce
> > a new relation "connected"? That is one or more ways and/or points could
> > be members implying that it is possible to navigate (perhaps directly)
> > between any of them. That would solve many of the problems for all
> > routers. This idea needs expansion: a tag on the relation would specify
> > the mode of transport, although I guess this would be mainly foot.
> > Likewise, some sort of seasonal tag would be useful. But most of those
> > already exist for ways, so I suppose those could be just be applied to
> > the relation as well.
> 
> Right now we represent this with ways, and they have well-defined
> semantics.  This seems comlicated and would need implementation in many
> data consumers.
> 
> > Just musing out loud. If it seems sensible, maybe a proposal on the
> > tagging list?
> 
> Feel free; I don't expect that to go well.

As far as I can see, the current discussion is just a special case of a
general problem with OSM. Consider a park with grass cover with open
access, or even just an area of, say open access moorland where it is
possible & sensible to walk anywhere. As far as I know, all the OSM
routers will fail to construct a route across such a place (presumably
a straight line) in the absence of an explicit way of some sort.

Adding a relation of the sort I suggested which would be verifiable and
objective would solve the problem. Is there a better solution?

Of course, routers would not pick this up immmediately.
As for mkgmap, I see that the Garmin firmware probably needs to support
it somehow. Are we sure that Garmin doesn't have something that can
interpolate between routing points? I guess that navit could probably
be extended to do this sort of thing.

ael