User talk:AM909

From OpenStreetMap Wiki
Jump to: navigation, search

Braided streets

I saw this edit on braided streets, and was wondering about this text you added:

It is a reasonable and well-used technique to bring dual carriageways back to a single point at their intersections to facilitate and simplify the mapping of control devices and turn restrictions

I'm curious about this, and was wondering where you've seen this. If you could link to some list or forum discussions or the wiki I'd appreciate it. I've never done this myself, but it seems like it would produce a somewhat ugly result (though yes, I know we don't tag for the renderer). However I suppose I can't object to it for any other reason that I can think of. -- Joshdoe 18:54, 7 December 2011 (UTC)

Having seen it done both this way, and by not bringing them together, and spending a lot of my time mapping intersections with turn restrictions, I found this way to be easier and less ambiguous. I saw existing examples of both, and remember at least one public discussion (sorry, no cite) on the subject that made it clear that the problem the web page was talking about was where the one-ways would cross over to the wrong side. From there, it progressed to include at least some of:
  • There was talk that the single intersection point makes it clear that it is a schematic intersection, not intending to reflect the individual lane combinations and actual signal positions. Of course, someone argued vehemently that we are not creating a schematic :(
  • Seeing 4 intersections (intersection of two dual carriageways), each with their own signal, can imply that this is the physical position of the signals and/or that they are independent (which they are not). If you place the signals in their actual physical position, or as nodes on the ways slightly away from the intersection, they become schematically incorrect, making routers do the wrong thing.
  • Multiple intersections presents the renderer with the need to combine signals and other traffic-control devices within a certain proximity to avoid clutter, or worse, not to do so. Either way, it's more work for the renderer. I don't think paying attention to this factor necessarily violates the principles behind not tagging for the renderer.
  • Multiple intersections creates ambiguity and duplication in the way turn restrictions are mapped (whether to use the short segments of the intersection or the intersection node as the via).
  • For a U-turn when using the multi-intersection approach, the "via" way must be the short section of the crossing way, forcing one to break that way into three pieces for this reason alone, and making it harder to maintain.
  • The correct place for a name change of a street is usually the center of the intersection with another street. Using the multi-intersection approach and ways for the via part of a turn restriction requires breaking ways into 4 segments and correctly naming and using them all, and can create clutter in attempted name rendering for those segments.
  • Explaining or understanding all of these nuances is much harder than the single-point intersection method. This is perhaps among the most important reasons overall.
  • Not that it matters, but I don't think the single-point intersection is ugly when rendered.
AM909 00:30, 14 December 2011 (UTC)