[Tagging] Roles of route members (was: Merging tagging scheme on wiki pages of Hiking, ...)

Kevin Kenny kevin.b.kenny at gmail.com
Sat Aug 17 16:37:52 UTC 2019


On Sat, Aug 17, 2019 at 9:44 AM Andy Townsend <ajt1047 at gmail.com> wrote:
> It does sound, however, that you don't have a concrete use-case at all - you have a view of how things "should" be, but this doesn't seem to be driven by a real-world requirement.  That's why I've been asking for specifics throughout this thread (and Richard has too).

There are some trails for which I wish eventually to produce map
books: a series of large-scale maps showing the route of the trail,
together with tables of linear distances along the trail (to obtain
distance between waypoints), and elevation profiles.  Of course, this
has the requirement that the "main stem" of the trail (I recognize
that there may be links, alternatives, etc.) needs to be _reducible_
to a coherent linear topology. (Note that I say _reducible_: I can
tolerate its not being actually _reduced_ in the database.) Answering
a question such as, "how far is it along the trail from the Route 28
highway crossing to the campsite at Salmon River?" needs such a
topology.

My operational definition when I'm mapping: "sorted well enough for
JOSM to show route continuity except where discontinuities exist in
the field or the mapping is incomplete, and for Waymarked Trails (or
similar software) to produce the elevation profile."

My operational definition when I'm consuming data: "the route segments
have endpoints that are sufficiently close to contiguous that a
reasonably simple, greedy algorithm can assemble a linear route"
(obviously, if there are route segments with specific roles in the
relation, that fact can imply a different topology). I shouldn't have
to resort to simulated annealing or machine-learning techniques to
assemble a route from a mess. If I get into needing to "approximate
longest Hamiltonian path", I'd be spending my time more effectively in
cleaning up the underlying data.

Ideally, I shouldn't have to resort to very much computation at all to
assemble a linear route, which is why OSRM flags all discontinuities.
But I don't complain when route segments are scramble. I fire up JOSM,
unscramble them, and leave it at that; I recognize that there are
editors that are less capable of maintaining a coherent route
relation, and that there are data consumers that require a stricter
one, so I just go in and repair it. I get rather more cross at way
ends that don't quite meet, because closing the tiny gaps requires
considerably more computation. I grumble considerably more at closing
gaps of a few metres.

I appreciate the work that the iD team has done so that I have to
repair these things much less often.

"Be strict in what you emit and permissive in what you accept" is a
reasonably good principle for data interchange in general.


-- 
73 de ke9tv/2, Kevin



More information about the Tagging mailing list