On Feb 11, 2008 7:45 AM, Bernd Raichle <[EMAIL PROTECTED]> wrote: > Hi, > > > on Monday, 11 February 2008 07:30:03 -0800, > Karl Newman <[EMAIL PROTECTED]> writes: > > On Feb 11, 2008 7:20 AM, Bernd Raichle <[EMAIL PROTECTED]> wrote: > [...] > > > > > > A few days ago I have started a new proposal for a "Segmented Tag", > > > which relates a set of tags to a directed or undirected part of a way > > > (I have called this part "segment" inspired by GDF's "Segmented > > > Attributes"). I have not found the time yet to finalize the proposal > > > adding some examples, nonetheless it can already be found in the OSM > > > Wiki (Relations/Proposed/Segmented Tags). > > > > > > Big +1 on this proposal. That's exactly what I've been thinking about > > lately. It's stupid to chop up nice long ways just because the speed > limit > > changes or the way happens to cross a bridge. > > Yes! > > Nonetheless we have to decide when to cut a long way into two or more > ways. > > - In the moment OSM is simple: > If atleast one tag value of a way changes, begin a new way. > > - Using the new relation "Segmented Tag", it gets complex: > You can have one long way with various tag values for different parts. > You can have even various "highway=" tag values for the same way! > > * What are the rules ... to start a new way? >
I think ways should be made as long as practical. The highway tag seems like a reasonable proxy for when to split the way. Generally the highway tag value doesn't change for long stretches. Also if the name of the way changes, it should probably be split. > > * What are the rules ... if the (ordinary) tag values of the way and > the tag values in the relation are different? (IMHO: most > specific rules, i.e., relation tags precede ordinary tags! :-) If we implement this proposal, we could deprecate tagging ways directly for certain tags (like tagging segments used to be discouraged), and flag it in the Validator for JOSM. > > * What are the rules ... if the tag values of two or more relations > for the same way part are different? (IMHO: inconsistency => error) > Hmm... In general, I would agree, but what about situations such as the "route" or "ref" tags? In the US, it's very common for multiple numbered highways to share the same roadway for some distance when they intersect, until they go their separate ways (pun intended) again. That would seem like you would want to use this relation structure to tag it, and that wouldn't be an error because the multiple "refs" are valid. For example, you'll have a state highway that joins the interstate, then runs along it for several miles until it departs again. That section of road is still considered to be the interstate and is known by the interstate name/number, but the roadway is often signed for both (with the "most important" shown on the top). You might want to call your proposal "section" or "sub-section" instead of "segment" to avoid confusion with the old segments from API <= 0.4. > Best wishes, > -bernd > > > PS: Btw, GDF has the capability of "Segmented Attributes", i.e., > attribute values which are related to a line feature segment only > (GDF line feature ~= OSM way). Nonetheless this capability of GDF > is _not used_ by the two major map suppliers. Thus if a line > feature attribute changes, the line feature is chopped and > attributes are assigned to whole line features only. I guess that doesn't surprise me too much. Their maps are intended for very narrow, specific purposes, and if they used the segmented attributes, they would need to pre-process the map data to chop them up anyway (which I expect we will need create a utility to do as well for the various data consumers). Karl
_______________________________________________ talk mailing list talk@openstreetmap.org http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk