[OSM-dev] Keepright layer conflicts
Harald Kleiner
e9625163 at gmx.at
Fri Aug 7 20:12:39 BST 2009
Hi Dirk
As Philip already said, I refer to the tunnel-wiki-page
(http://wiki.openstreetmap.org/wiki/Tunnel#How_to_Map). It says that you
should have at least a short non-tunnel-way between a tunnel and the
next junction.
From this I conclude that the only valid method of changing layers is
exactly two ways joining in a node end to end, like this:
way A, layer 0 way B, layer 1
-----------------------*----------------------------
According to this rule, this is obviously wrong:
[two ways connected together at intermediate nodes while on different
layers]
way A, layer 0 |
|
| way B, layer 1
--------*------------------------------------
|
|
this is IMHO also not OK (but nevertheless common practice):
[a bridge ending at a junction]
way A, layer 0 |
|
| way B, layer 1
*------------------------------------
|
|
There are many junctions of the third form, I admit. That's why I
separated the second from the third form by different error types. So if
you don't like to change the way you draw bridges/tunnels you can simply
turn off the "not so obvious" layer conflicts errors.
Best regards,
Harald
> Hi,
>
> Am I the only one with a problem how keepright handles layer conflicts?
> In my view ways connecting at a single node are connected, even if they
> run on different layers. That is the only way how to connect layers at
> all, so the check should at most hint about a possible error when ways
> of completely differend kinds (e.g. highway and waterway) join with
> different layers and the shared node is a middle node for both ways.
> especially if the node is an end (or starting) node for any of the
> joining ways this should not be considered an error (for that way) in
> any case. (where three ways meet and only one uses the node as an end
> node this actually might be an accidental connection made by the editor
> software).
>
> So please leave the check out as it now reports abundantly many false
> positive and only few legitimate results.
>
More information about the dev
mailing list