Oof. In that case I'm going to have to revert a *lot* of my recent changesets. Still, all a good learning experience.

(For the rest of this post, I'm kind of thinking out loud - don't feel under any obligation to answer my annoying questions, although I'd be most grateful if you do.)

When you say the imported parish boundary data are correct: the real boundaries are (for the most part) smooth curves, and the imported data are some polygonal approximation to those, right? And it looks like the granularity of the polygonal approximation is about 15 metres: in some cases, that approximation is in itself enough to qualitatively mess up the correspondence with physical features. Is there anything in particular we're supposed to do about that?

It's tempting to respond to your point by saying "in that case, where a parish boundary is known to coincide with a physical feature and they're currently mismatched in OSM, we should displace the physical feature to match the parish boundary". But the reason the parish boundaries are known to coincide with physical features is that the County Council has presented its own (definitive) map of the parish boundaries overlaid on an OS map that includes the physical features - so this would be a slightly indirect way of copying the physical features from an OS map, which I believe is a big no-no, isn't it?

Something I've encountered quite frequently is that a parish boundary coincides with one bank of a river or stream or ditch, but that river or stream or ditch is implemented as a zero-width way in OSM. I was never very satisfied with the way I was handling that, but in the light of what you've told me, the way I was handling it was very wrong. Is there a policy for that circumstance?

I was about to say something based on the premise "I need to learn to revert changesets", but from what you've just done, it looks like the method of reverting a changeset is to create a new changeset that does the opposite things, in which case I've caused you a bigger headache than I imagined, so many apologies.

I was in two, or maybe three, minds about this. Legally, English parish councils (admin_level 10) derive their existence, powers, and territorial extent directly from legislation by central government (admin_level 2). Anyway, it looks like at least part of what I've done is a big policy no-no and needs reverting.

(In fact it sounds like you're talking about me adding local government admin boundary relations as subareas of the UK admin boundary relation, right?)

I think I've only added the one example of the non-standard "townhall" role in a boundary relation (precisely because I wasn't sure it was acceptable practice and was waiting for review before adding any more). But maybe you're talking about some wider set of the things I've done?

Network Rail risk-assesses the #20590608/#2187758447 crossing as "C2", meaning very high risk both to individual users and to the community as a whole.

(Also, according to abcrailwayguide.co.uk, that crossing is not called "Della". The one called "Della" is about 400 metres further southeast.)

The level crossing marked as nodes #20590608 and #2187758447 has user-operated gates with no lights/bells/automated locks. Frankly, it's terrifying. I wonder if we should add some kind of hazard tag to it.

There are several ways in this area (#1288228947, #1288228945, #303999843, #1125885582, #1125885654, and #532977208) which are tagged with prow_ref "Ugborough footpath 30" even though they're beyond the ends of Ugborough footpath 30 as it appears on the interactive "working copy" of the definitive map on the Devon County Council website (and in some cases outside Ugborough parish territory). Not sure if that indicates something wrong with the tagging on here or something wrong with the interactive version of the definitive map...?

Come to think of it, why is there a 30-metre contour in the grounds of Sawston Hall? There's no way that area's within 1 metre of the peak elevation of Huckeridge Hill is there?

Taking a look on OpenTopoMap, the way-354596796 stream doesn't have consistent "uphill" and "downhill" directions - e.g. it crosses the 30-metre contour four times, twice in each direction. (And even if it did have consistent "uphill" and "downhill" directions, there's no guarantee that the stream bed slopes the same way as the surrounding land.)

I mean, I think the change you've just made is correct, and it flows east to west from its junction with way/354596799 to its junction with the river Granta, but it's maddeningly difficult to be sure.

Thanks. Done. Did you see that I asked the same question about another nearby stream (way/354596796)?

How was the flow direction in the stream (way/354596796) determined, please? The map accompanying the inclosure award for Sawston labels it as a "drain", which I'd instinctively interpret as meaning it flows towards its junction with the River Granta, not away. Sadly, it's currently dry (at least at its crossing points with Sawston High Street and Sawston footpath 9), so I can't check by throwing a leaf in and seeing which way it floats.

How was the flow direction in the stream (way/642898893) determined, please? The map accompanying the inclosure award for Sawston labels it as a "drain", which I'd instinctively interpret as meaning it flows towards its junction with the River Granta, not away. Sadly, there's no public access to any part of it to go and check.

Oh yes, and at the end, for this specific changeset: I discover that the ditch/track crossing is a bridge rather than a culvert - the probable OS basemap data is my sole source for that.

Thanks, Jez. The process for this feature was as follows:

- I know there's a ditch called "drain number 3" in the area, and its approximate course, from the map accompanying the inclosure award for Sawston (1809, so definitely out of copyright).
- I confirm that the ditch still exists, and still follows the same approximate course, from the map on the CCC website (I think that probably is OS basemap data, although I can't be sure because I don't have access to the unadulterated OS basemap.)
- I mark the exact course by drawing a line following the linear feature that's visible in the satellite imagery provided on iD.

What do you think? Is that an acceptable procedure?