On 05/03/2016 03:09 PM, OSM Volunteer stevea wrote:
In the USA, partly because we are such a geographically large part of the North American continent and partly because each of our fifty states is sovereign, I find that breaking apart very large relations so they are across a single state at a time (then perhaps these are collected into a super-relation) is often (though not always) a sensible approach. It is part size (large relations with vast numbers of members are unwieldy), it is part “what sort of an entity is this politically?"

For example, there is a note in OSM’s Amtrak wiki page on the route=train relation for the California Zephyr: "The relation is said to be so big it is hard to work with.” That is something we might take to heart and break apart the relation into statewide components. I haven’t done that, but somebody might, after considering that it makes editing easier, and that state-at-a-time is a good way to do this. Even a simple web browser request to display this relation results in "Sorry, the data for the relation with the id 905830, took too long to retrieve." The practicality of potentially better avoiding edit conflicts has been mentioned, and is also true.

Breaking apart the AT into separate relations - ideally with a superrelation joining them - would be sensible, I think, but be careful about the assumption about state lines. The AT literally spends a good many miles with the hiker having one foot in North Carolina and the other in Tennessee - the ridge that it follows is the state line.

We also, I think, need to put some more thought simply into the support of large relations. I've recently found that even the New York Long Path (only a fifth the length of the AT) crashes JOSM (I haven't yet diagnosed the problem) and wound up editing in Meerkartor instead. Trails, highways, rivers, railroads, we have a good many places where things reasonably and predictably break down into thousands of parts over thousands of km, and I don't think we yet have a unified theory of how to handle them.

--
73 de ke9tv/2, Kevin


_______________________________________________
Talk-us mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/talk-us

Reply via email to