[Talk-us] Relation roles

[Chris Lawrence]

On Wed, Jun 29, 2011 at 5:52 PM, Paul Johnson <baloo at ursamundi.org> wrote:
> One should hope the software can figure that out based on the overall
> orientation of the relation.  I'm in the forward/backward crowd myself,
> as it works well for routes that are only sometimes divided.  For routes
> that are always divided, rather than making it more difficult for the
> tools to work with, a super relation containing relations for each
> direction works well.

The problem is that there are routes that change directions (I-69, US
83 in south Texas) and even some that violate geography ("east" and
"west" on much of I-26 in East Tennessee are 180 degrees off, and
don't even make sense when you account for the 45 degree rotation of
routes in that area).

Having said that a relation for each direction, even for undivided
routes, is probably superior just simply because it allows turn
directions in navigation apps to be correct by reference to which
relation is in the direction of travel.  (e.g. if you're "backward" on
a way you can find the relations where the way has the backward role
and use those ones as the directional indicators).  That this also
helps the relation checkers and sorters (which could after all be
fixed to cope with directional roles - and sorting in JOSM copes to
some extent with directional roles already if all the ways are loaded)
is a bonus.


Chris