<html><head></head><body>Hi Christoph, my suggestion was to clearly separate the subject of water flow from the subject of routing. Whether roads are mostly bidirectional or not is irrelevant I think, as routers have to be able to handle one-way roads anyway. If I understand it right, edges in routing graphs are often one-way anyway as the characteristics from A to B may be different from B to A.<br>
If you are interested in hydrodynamics on a large scale, like predicting the route an unpowered floating object would take, that is entirely different to the use of a routing engine to suggest a route to the skipper of a boat.<br>
Simply adding a way from one side of a lake to the other to stop some QA program complaining is bordering on tagging for the renderer...<br>
--colin<br><br><div class="gmail_quote">On 28 July 2015 11:17:00 CEST, Christoph Hormann <chris_hormann@gmx.de> wrote:<blockquote class="gmail_quote" style="margin: 0pt 0pt 0pt 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">
<pre class="k9mail">On Tuesday 28 July 2015, Colin Smale wrote:<br /><blockquote class="gmail_quote" style="margin: 0pt 0pt 1ex 0.8ex; border-left: 1px solid #729fcf; padding-left: 1ex;"> If we can separate the flow direction discussion from the routing,<br /> the latter becomes a more generic "routing through areas" problem<br /> which has been discussed before in the context of pedestrian routing.<br /></blockquote><br />Water flow structure is not only about flow direction of individual <br />segments, it is also about connectivity - hence the routable network.<br /><br />And yes, you can try to take polygons into account to determine <br />waterflow - you have to at the moment since missing line mapping is <br />just too widespread. But for non-trivial polygons (i.e. ones with <br />holes or areas represented by multiple polygons) this is generally <br />ambiguous and it is hard to analyze as well.<br /><br />The analogy between water flow analysis and traffic routing is <br
/>misleading here since traffic routes are mostly bidirectional. If you <br />imagine a road network exclusively built from oneway roads you can <br />immediately see that having parts of that network represented as <br />polygons will make routing difficult.<br /><br />And - this is even more important - it also makes it difficult to spot <br />errors in mapping. If you have a full line mapping of a river network <br />it is very easy to identify problems locally for QA tools or validators <br />in editors - just like in case of the coastline. If you rely on <br />polygon features creating connectivity in the waterflow network you <br />need to analyze it in full before you are able to spot where mapping is <br />broken.<br /></pre></blockquote></div></body></html>