[OSM-talk] A post box called Breuningsweiler

[David Earl]

Simon said:
> So, I'm left with a dilemma. Do I move the centre of the village just
> because the it's now complemented by a mini roundabout,
> or do I leave out the rather useful fact that there's a mini roundabout
> where you may need to turn off the A271 to reach one of the other
> streets (and that's just after a new pedestrian crossing)?

A few metres either way on the "centre" isn't going to make any difference
to a routing algorithm, which will, in any case (as this is the majority)
have to take account of places which aren't connected to ways at all (so
it's no use saying 'putting it on a way will help routing algorithms' -
which it might - unless everyone changes the way they do it).

So why not just put a node alongside (a metre or two away from) the junction
and put the place there? Either in he highway, or preferably a distinct
node?

Actually, putting the node at the centre may not be the best approach
anyway. I would hope people using a routing algorithm would ask to go to a
street in a place, not some nominal and usually arbitrary centre of the
place (and as I said yesterday, the centre of very many places is not
accessible by car any more); but more important at the moment is that this
is where the imformation density is likely to be highest and the name
rendering on the map may well conflict with the other information. Of
course, renderers should be cleverer about how they place text, but it is a
very hard problem, so we can expect the name will go where the node for the
place is, at least for the forseeable future. So a slight offset into a
blank area near the middle may be a more judicious choice.

If we end up with place names on nodes shared with other features, then we
must have other more complicated rules uch as a hierarchy of tag types such
that if there is a name, it applies only to the object higher in the
hierarchy; or say things like mini roundabouts don't have names (sounds
reasonable at first, until we realise that junctions very often are known by
a name - there are several in Cambridge - 'Murkett's Corner' for example is
a junction characterised by a set of traffic lights).

It seems to me to be a much simpler rule that we say anything can have a
name, but you don't introduce ambiguity in what the name can apply to.

All the features on the map are approximations to some extent, and places
are really are areas not points, so exactly where in space the place point
goes is really not that important. The consistency of information to make it
processable seems much more important to me.

David