[OSM-talk] turn restriction relations: via

[Tobias Knerr]

I've been implementing turn restriction support for some (still
unpublished) software recently -- and, of course, have also added every
restriction I could find to OSM. While the concept of "restriction"
relations generally works well, there is one thing in particular I'm not
really happy with: the effort that is required to handle via members.

Can/should I assume
* that there will always be exactly one finite sequence per restriction
that includes all via members?
* that via ways will be split to connect with each other and with
from/to only at start/end nodes (this is recommended for from/to, but
not mentioned anywhere for via ways)?
* that a relation using via nodes will always contain all nodes as via
nodes one would travel along
* that a restriction will include either via ways or via nodes, but not
both?

Making these assumptions wouldn't prevent the description of any
real-life-situation. It would, however, make evaluating restrictions (as
well as understanding the effects of a restriction as a mapper) more easy.

I also wondered whether it is really necessary to allow both (multiple)
nodes and ways as via members, as each possibility alone can perfectly
handle all situations and there is no significant difference in mapping
effort either. So why don't we restrict the via role to ways (except
maybe for the popular single-via-node case)?

Tobias Knerr

PS: I'm perfectly aware that there is no authoritative documentation,
that I can implement and map whatever I want etc. I'm just interested in
other people's opinions.