|
On 2013-02-23 20:02, Jo wrote :
It seems that you would like a specific role, which
you can add to 2 members of a route relation (I'd add it to the
two ways around your imaginary gap).
If you do it that way, you don't need a non-existing member. And
you don't need to add nodes to a relation which consists of ways.
Yes I want to add a new specific role but it cannot apply to the two
ways around the gap for the simple reason that these ways may
already have another role. I have already discarded the dummy way.
It turns out from my text that, in reality, I'm filling the gap with
a GPX trace (we might use "track_point").
That means that we must add nodes and just nodes (to the relation).
If the GPX trace contains only two nodes, the near nodes of "the two
ways around" must be duplicated and it's an easy task for validation
software to check that configuration (as if there was a way using
the two nodes).
If we were allowing such "GPX traces" of more than two nodes, we
would, in addition, have to invent sort of dummy nodes with at least
sort of GPX=yes tags so that they're not yelled at tagless isolated
ones and that we know what they are. Validation is the same except
that there's absolutely no clue to validate the extra nodes.
It should be noticed that, in converting a GPX trace to a route,
such gaps are the leftover, the GPX pieces that could not be
converted. It seems to be a logical thing to do to keep them as GPX
(simplified, of course) ... until Osmose warns that a new road has
been built ;-)
This doesn't just have implications for the validator, but it also
might involve changing the code which sorts the member ways.
?
On 2013-02-23 19:53, A.Pirard.Papou
wrote :
On 2013-02-22 12:10, Janko Mihelić
wrote :
I'm not entirely sure I understood your question,
but you shouldn't map non-ways. Routers could be developed
that route through non-ways, if there is no cliff or something
else in the way. A router could route along the contour lines,
to make the hike through forest easier. But if there is no
path, don't map it.
On 2013-02-22 14:05, Volker Schmidt
wrote :
It happens often on mountain hiking routes. You
have a signpost with the red-white sign of the Alpine Club
that indicates the direction that you have to take across a
meadow, for example. On the other side you have to find a
corresponding sign. In between there may not be any visible
path. In that case I would happily put a highway=path with
surface=grass as a straight line across the meadow.
On 2013-02-23 12:56, Martin
Koppenhoefer wrote :
maybe add the key "informal"=yes to the path? I do
this for "spontaneous" ways and it is also documented in the
wiki: http://wiki.openstreetmap.org/wiki/Key:informal
And the other suggestions, many thanks, sorry for not listing
them all.
I'm looking for a general feature, not only a solution to my
particular problem.
A non-way is not the best word to describe my idea and I also do
not feel comfortable with it.
It's sort of a "secret [winding] little passage" that one must
follow on demand.
So, more than "informal=yes" (which I don't understand well), it
would be a straight "exists=no".
How could it be mapped, sort of dotted line, so that the human
understands that he may follow a route for which there's no path
under the conditions otherwise described (no cars in a meadow)?
But how could the automated router know if it must or not follow
that secret passage?
Mind boggling, it needs more information.
And these thoughts led to the following reasoning...
In making a route (the relation), we are actually not mapping
something (creating new map objects). We are relating existing
objects of the map to be highlighted to show, well, a route to
follow (other relations similar).
And it may, for many various reasons of which you found more,
happen to be NO objects in the map to highlight and follow. So,
this problem is just, within the queue, aka file, of members
making up the route, to indicate somehow: this gap is not a
mistake ("page intentionally left blank", JOSM don't complain):
it means that you just must manage to go from here to there the
best way you see fit, para-gliders included
(1).
The first idea was to fill the gap with a dummy, but the second
thought is that we simply could use the end nodes of the two
ways the gap is striding to do so.
One node, repeated next to the way it belongs to, would have
role gap_start, the other one gap_end.
Or jump_start, jump_end
(1).
No dummies needed.
Human routers (mapping a hike) just assemble these special
instructions among the members.
Automated routers are driven by a human who simply breaks the
route in segments (making "via" points), one of which uses no
car, bike or pilgrim type but that funny little flying bird as
the segment routing type. By definition of the crow segment, the
router makes it of only two gap-start and gap-end nodes (it may
use more nodes and, magically, we reinvent the GPS trace (we
might use track_point instead of gap_*, but that
would lessen the possibility to detect routes broken by less
capable editors).
I think it's a rather simple, best value for money, addition to
the OSM tags I let you discuss.
To end my practical story, not only do the hike instructions
loosely say that the hike starts and ends in the parking place
(which is obviously the car segment of the hike!) but the bird
segment starts wandering north in a drunkard fashion where there
is no path, even breaking its way through the limit of an
alleged cemetery.
I simply started on the road alongside the parking and cheated
my way trough a small street detour.
They call that a walworkaround ;-)
Cheers,
(1) Yet another real case of possible exists=no routes coming to
my mind errr... BTW.
Ski routes too. Endless.
2013/2/21 A.Pirard.Papou <[email protected]>
Hello world,
It can happen for a hiking route, maybe others, to go
across a non-way. One may for example get people across
some land without a path or officially start and end a
hike in the middle of a parking lot.
What must we do:
- create a pseudo way and what are the tags?
- more likely, leave a gap in the route relation,
filled with some element saying "fly to connect"?
The crow may be supposed to fly loosely following the
roads too if router software is unable to make a
correct route or simply if the user insists on being a
crow. This is not a mapping issue, but the solution
can be the same if the router builds the same relation
as ours as the output of its result.
I suggested several sites to add a flying bird to car,
bike and man to be chosen independently per segment.
This (unable), in addition to map bugs, is the case when
using say the Google router with an OSM map display.
e.g. openrunner.com's
doc says to use Cloudmade router but soft only provides
Google's on OSM.
(You'd do something nice reporting this bug).
I only found the following in close relation with this.
In two parts (yes, sometimes the gnus have to fly too
;-))
http://lists.openstreetmap.org/pipermail/talk/2010-November/055088.html
http://lists.openstreetmap.org/pipermail/talk/2010-December/055121.html
Cheers,
|
