Reporter: david[at]frankieandshadow.com
[Submitted to the original trac issue database at 6.27pm, Monday, 21st November 2011]

In the centre of this map
http://www.openstreetmap.org/?lat=52.21402&lon=0.09524&zoom=15&layers=M
you can see "Gardener's Cottage". This is a completely insignificant little building in the middle of a major site of the University of Cambridge called "Madingley Rise". But you have to zoom right in to max before you see that important caption:
http://www.openstreetmap.org/?lat=52.214342&lon=0.094425&zoom=18&layers=M

This is just an example. There are numerous other examples even in Cambridge (the ones I'm most concerned about are University sites and colleges not being rendered in favour of some comparatively trivial building or other feature within the site or college).

I suspect the caption collision is preferring the trivial over the significant here. Is it random which it chooses? I think it should use some metric: e.g. the larger of the two areas, larger of the bounding boxes of the two areas, the one with most constituent segments, or total segment length or something like that to give a heuristic for which is probably the more important caption.

I imagine you would favour city over suburb when those two clash, but there you've got some clues in the tagging. But the principle is the same "larger is more important" for some definition of larger.