<div dir="ltr">My opinion on this would be as follows...<div><br></div><div>junction=roundabout being anything that's roundaboutlike, where entering traffic yields.</div><div>junction=mini_roundabout being the same, that does NOT have a hard median in the middle, and is just painted on (rare enough in North America that short of photographic evidence to the contrary, I'm very sure NASA has had more black astronauts in it's history than there are mini-roundabouts existing on the entire North American continent).</div><div><br></div><div>I've tagged dogbones as roundabouts, due to their roundabout-like nature, just abnormal shape (they are functionally a traditional roundabout joining a diamond interchange to a surface route via an over or underpass). Weird cases that I'm not sure on would be partial dogbones, where a three-way roundabout cannot be completely traversed and from which the surface street doesn't continue and is controlled by a red light that only illuminates if traffic from the motorway backs up (AFAICT, a one-off Maryland case), traffic circles (roundabout, but priority controlled entirely by traffic signals, those on the circle may have to stop for the light), and some of the cases that have come up in this thread (I'm unaware of even a North American equivalent term for these situations since they're literally not a thing that I've ever encountered where traffic on a circle might not have the right of way or a traffic light).</div><div><br></div><div>I recommend against using rotary as a term for anything ever, since for most of North America, it means the same as beltway, except New Jersey and maybe Delaware, where it's a generalized term for both traffic circles and roundabouts.</div><div><br></div><div>I'm all for additional junction=* definitions. Could even go relational with this since some roundabouts, particularly where multiple highways/bus routes/bicycle routes/etc overlap (even in countries like the UK where routes "never" overlap) exist. And a relational junction would be handy for particularly obnoxious junctions, like, for example, where a Texas-style expressway (four-carriageway!) with an additional separate carpool-only dual-carriageway crosses a dual-carriageway, and the six-carriageway has Texas turnarounds to completely bypass the SPUI for the outer carriageways and the crossing dual-carriageway. <a href="http://www.openstreetmap.org/#map=19/33.03758/-96.70421&layers=N">http://www.openstreetmap.org/#map=19/33.03758/-96.70421&layers=N</a> I wish I could say this was even *remotely* the hardest junction in Texas to map. It's not. Just look where I 635 meets US 75...then scroll east....for around a mile the freeway is (OTOH) 9 carriageways...</div></div><div class="gmail_extra"><br><div class="gmail_quote">On Fri, Nov 18, 2016 at 1:14 PM, Tom Pfeifer <span dir="ltr"><<a href="mailto:t.pfeifer@computer.org" target="_blank">t.pfeifer@computer.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">We had plenty of discussion over three days, unfortunately without a resolution. We were discussing specific roundabouts, and what the best router announcements would be in various cases. We did not decide about a value.<br>
<br>
Currently, OSM defines junction=roundabout with two conditions fulfilled, i.e. going round a non-traversable island AND circular traffic has right of way (when potential traffic lights are off)<br>
<br>
Nobody objected the idea to assign a 'junction=' tag to installations that fail the OSM definition above, in particular with different right-of-way situations (all ore some incoming roads have priority). The only objection was against "not-something" troll tagging.<br>
<br>
So far we had on the table, with taginfo in brackets:<br>
<br>
1 junction=circular (0)<br>
2 junction=circular_crossing (0)<br>
3 junction=circular_junction (0)<br>
4 junction=traffic_circle (35)<br>
<br>
For comparison, junction=roundabout (429490)<br>
<br>
My strong preference goes for 1 "junction=circular", as it is concise, has not a biased meaning towards the roundabout-definition, and is highly descriptive, namely a junction that is circular. Thanks Martin for proposing.<br>
<br>
2 does not fit so well as we have no crossing. 3 duplicates the term 'junction' in the value, we have that in the key already.<br>
4 'junction=traffic_circle' is the N-Amer synonym for BrE roundabout, thus might be ambiguous. The 35 current uses need to be ground-checked what they label.<br>
<br>
The definition for the new value needs to restrict the size a bit, e.g. not having other roads inside the circular junction. Thus we would exclude e.g. the motorway ring around a city.<br>
<br>
Please note that we are just discussing the _tagging_ here. The routers and renderers can still use it as they like, and I still do not like announcing exit numbers on a large roundabout/circular junction.<span class="im HOEnZb"><br>
<br>
On 07.11.2016 15:30, Daniel Hofmann wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
4/ Not-a-Roundabout (what this post is about)<br>
There are situations where one of the entering road has right of way,<br>
which disqualifies the scenario for being classified as a roundabout.<br>
The Wiki has a section on these Not-a-Roundabouts:<br>
</blockquote>
<br>
<br>
<br></span><div class="HOEnZb"><div class="h5">
______________________________<wbr>_________________<br>
Tagging mailing list<br>
<a href="mailto:Tagging@openstreetmap.org" target="_blank">Tagging@openstreetmap.org</a><br>
<a href="https://lists.openstreetmap.org/listinfo/tagging" rel="noreferrer" target="_blank">https://lists.openstreetmap.or<wbr>g/listinfo/tagging</a><br>
</div></div></blockquote></div><br></div>