True , although that is implied by the size of mini-roundabouts and the
time taken for traffic to reach you.
Traffic does not generally slow to the speed needed to go around the
paint.
There is also the rule that you shouldn’t use mini-roundabouts to make
U-turns, which routers would need to be aware of.
Basically we have been using this method forever, it works and as they
say 'if it ain’t broke, don’t fix it".
Casey_boy and SekeRob have just shared some images of it “working”.
I believe if it worked, we wouldn’t have this discussion returning 
I think if this is causing people to “bowtie” intersections we need to mini roundabouts on circular ways (probably with a junction tag).
As a bonus it might mean we get magic roundabout mapping that reflects the usage.
THis Magic Roundabout, does it have two-way traffic between the mini-roundabouts?
Most definitevely.
From what I can tell on the aerial view and from people discussing it (I haven’t driven it myself):
It’s a two way ring of road with a mini roundabout at the entrances of the proper roads. The issue from an OSM perspective is that it’s a dual carriageway. If it was a single carriageway it would be fairly simple to draw (it would resemble the sign), but to have a separate way for each direction someone has added fictional “radial” roads going from the centre of each mini roundabout to the joining parts of the dual carriageways.
If only it was a single carriageway:
Took a peek. That someone has made it very very complicated. Though I admit that the inner circle of the larger circular junction could be seen as a large bypass, not entering the miniroundabouts untilthe driver takes a turn onto one of them.
I think in your picture the circular sections could simply be two-way. Then it’s just a series of mini roundabouts connected by a two-way road which happens to turn onto itself.
To map the circular as two separate cariageways (a valid choice, because of the physical separation islands), I think I would choose to map the mini-roundabouts as roundabouts, just smaller than usual. I.e. a oneway ring, and each approach and exit lane lands on its own junction node on the circle instead of a central node. Takes much more work, but would route ok I think, and the rendering would be much more realistic.
Does look like a bowtie indeed 
and thought of the logically following… junction=bowtie_roundabout, except physically not there, a complete misrepresentation. Leaning to untie this one and restore the prior art.
(Surely there must have been competitions to see who can do the 5 in 1 ‘magic’ circuit the quickest.)
I’ve borrowed the bowtie description from somewhere, just not sure where. I thought it was the Mapbox suggestions for mapping intersections but they compare dual carriageways done in this manner to sausages instead. I think it’s geometry we should avoid if possible as it doesn’t really represent the layout or how the vehicles move in a meaningful way.
I you look at it as 5 small roundabouts with short but regular ways in between, it doesn’t seem that comlicted to me. The whole thing is not a roundabout for OSM, because roundabouts are implicitly oneway traffic. Complicated is: mixing a separate way approach with a lanes approach.
I think the separate way approach (including a circlular way on the roundabouts instead of a centre point) would give the most realistic result, resembling the aerial view and the announcement signs. I wouldn’t mind having a go, there’s just one thing: driving on the left is very counter-intuitive for me, I am bound to make stupid errors because of that. The effect on navigation instructions, I can not predict. I’m wondering if the instructions and announcement make any sense as it is?
I think this is all essentially correct, the issue with mapping it that way is that they are mini-roundabouts and we currently map those as single nodes. This is why I mentioned it on this thread as something that could be improved by us allowing circular ways for mini roundabouts. As far as editing it, I’ve thought about it a few times and even drafted something once, but the mini roundabout as way issue, the intimidating number of relations going through it and the thought that I might annoy a mapper who actually uses it have all put me off.
I’m unsure whether junction=roundabout+roundabout=mini_roundabout or junction=mini_roundabout is better. The first pair of tags would allow existing routers to get things right 90% of the time, but may lead to situations where an unaware router encourages a U-turn on the mini roundabout which is discouraged (at least in the UK). The latter breaks things for routers who don’t understand the tag, but avoids diluting the meaning of the main roundabout tag.
In places where there’s a legal difference between mini and non-mini roundabouts designation=mini_roundabout might be appropriate if we used the roundabout=mini_roundabout method.
MR’s are UK only I think?
So far, I have only compared the legal situation of GB and Germany. The only real legal difference between a roundabout and a mini-roundabout in GB is the traffic sign and the white inner circle that identify a mini-roundabout as such. Normal roundabouts in the UK usually have normal right of way signs and/or road markings.
Apart from that, the same road traffic rules apply in the UK as for normal roundabouts, they are only listed separately and explicitly because more specific things have to be observed in relation to the often really small diameter in the UK. The mini-roundabouts in the Road Traffic Regulations are therefore only a sub-item in the section on roundabouts.
Often - resolutely - put forward arguments such as:
are, in my opinion, not reflected in the British road traffic regulations in such absolute terms.
All vehicles MUST pass round the central markings except large vehicles which are physically incapable of doing so. Remember, there is less space to manoeuvre and less time to signal. Avoid making U-turns at mini-roundabouts. Beware of others doing this.
The “fact” that british drivers would expect different turning instructions has already been described differently here by a british driver in the UK:
Note that historically, when the British invented the mini-roundabout, these mini-roundabouts were really mostly very small and replaced normal equal-ranking intersections. Mini-roundabouts with a central island that can be driven over have now triumphed all over the world, although outside the UK this may be differentiated in traffic regulations.
Even if the really small mini-roundabouts still predominate today, the building regulations have also changed in GB in the meantime. Mini-roundabouts (newly built) in GB can have an outer diameter of up to 28 m. This is actually larger than roundabouts. This is actually larger than roundabouts in Germany, which are considered “mini-roundabouts” in construction terms.
And these larger mini-roundabouts already exist!
Therefore, I think it may also be important to the British to map larger mini-roundabouts in OSM as circles and still not lose the important information that the inner circle can be driven over.
translated by Deepl
I think the REAL problem here is that the description of the both just has the traversable center as the difference. So mappers wrongly assume that everything with a traverseable center is a mini_roundabout which it is not.
We need to fix these issues in the wiki.
Flo
On mapping the junction as a node or as a circle: I ask myself, if a split road (two carriageways) crosses another split road, where the junction area in fact is one undivided surface, would I put a junction node in the centre and connect all the carriageways to it?
They certainly exist in France.
Maybe we need a roundabout tool that could replace a mini-roundabout with roundabout (and the other way round), keeping the relations intact?
After all, these relations have some entry and some exit from junction, so it should be possible to connect them…
That would be nice, but in JOSM it’s not that hard to correct the relations. I think it’s more important to agree on when to do it and when not to do it. Maybe the presence of flare islands coud be decisive: if you don’t split at least one of the ways, there is IMO no point in using a junction=roundabout circle instead of a highway=mini_roundabout node.
I tried to make sense of the French rules for rondpoints et carrefours à sens giratoire, but failed miserably…
Looked like only carrefours à sens giratoire are roundabouts for OSM, and there is no particular size limit for these roundabouts. While rondpoints are not roundabouts for OSM because there is no priority for traffic on the circle. So they are not OSM mini_roundabouts either. Correct?
PS I looked at this one which has no roundabout sign at all, which means it is neither a rondpoint nor a carrefour à sens giratoire, I think? Just a slight traffic calming bump on the road.
PS2 Ah excusez… I had another look. At quite a distance from the junction, there are warning signs.
But these things are independent. The tool would, first of all, identify all relations needing care after changing. Then fix relations automatically.
Redefining when to use which tagging has less impact on thatbhypothetical tool than deciding on actual tagging (do we keep current concept or define some common tagging for both?)
OTOH, your question might need answer before or paralelly with decision on tagging…
#/media/File:Magic_Roundabout_Schild_db.jpg){kind=link}