It turns out, vehicles can do a (special) u-turn to turn towards the Prince Edward Road East bridge. afaik there are basically no intersections in HK where a non-dedicated u-turn is required to get to a destination.

Earlier I did not know/notice this, and put a note in this area ("Dakota Drive should not get =secondary"). Still, it can't possibly be "highway=secondary", which causes confusion; the road does not "continue" like this. If it were "=secondary_link" it would be much clearer.

Correct observation on layering. Two points:

- "free-hanging" layers allow for convenient tagging, touch wood we want to map everything from the ground floor up till 17/F, then we can just do e.g. "layer=1" because everything at 17/F is already "layer=17" and doesn't interfere our work
- iD editor; it currently cannot resolve intersection warnings without "level+layer", or maybe I am misunderstanding something

You are correct. "Traditional methods" aka GPS observation doesn't work here.

These are all reasonable deductions on top of "good enough" existing data, sometimes combined with vague irl experiences (in case I did travel along the segments); techniques:

- station/platform size sanity check (use iD editor measurement feature; combine with irl vibes "it can't be this small/large")
- railway basic requirements (curves are almost always circular arcs of some form to not stretch/compress axles)
- railway speed requirements (medium speed rail does not permit sudden curves)
- tunnels avoid buildings (boring would otherwise intersect building foundations, esp tall buildings + shallow tunnels, and we don't want that)
- buildings avoid tunnels (quite obvious; "railway protection" etc, foundations must be shallow to not intersect tunnels)

But on top of that, when in doubt, use a bit of art sense. That's why it's usually called "calibrate/beautify", in case you noticed.

Good point; I notice rural addressing can be confusing when they do "Number XX, Village".

Improved via changeset/175882185

Re 1, in case you missed it, from the same quoted discussion:

> that would be for another changeset and for separate survey/lookup work.

Re 2: I will not act until there is an obvious and uncontested replacement.

Re 1, also see discussion of changeset/174721133

Re 2, it seems we do not have clear consensus of what values to use.

Improved with changeset/175799549

PS maybe bookmark eg https://osmlab.github.io/osm-deep-history/ for future convenience

I vaguely remember in the past the iD editor had "Vehicles" instead of "Motor vehicles" with the intent to describe "car" access, which would bump the `vehicle` tag.

But still, to strictly follow definitions, I can further edit it to be `vehicle=no`.

The "contraflow bus lane" can be simplified as "this road will only lead into the bus terminus". I view it as an extension of the bus terminus.

Normally the "this is bus terminus, do not enter" signage is placed at the proper entrance, but for convenience and reminder purposes, it's placed earlier.

It's like you wouldn't need a "no left-turn" OSM restriction when the left side is oneway coming out.

Ideally routers should learn to recognize "impossible dead ends" and exclude them from general navigation.

I prefer things to be simple.

No idea, but at least should not be in the name field. I do not specialize in addressing.

`motor_vehicle=no` was added in the past, not me.

Old style is `vehicle=*`, new style is `motor_vehicle=*`. This is mostly iD editor changes, but also consider the fact that "vehicle" and "motor vehicle" in practice are largely equivalent (at least in HK).

Feel free to improve this if have more info.

I couldn't determine/notice what this is irl, but then I saw this from aerial imagery.

Might be too late to type this, but some "rest areas" might be "misunderstood tagging" and should be normal parks instead.

Afaik the whole "footbridges in Tsuen Wan" network itself *is* a construction project that the district council pushed for in the past decade. imo it deserves a relation of some sort.

This is not a complaint, but more like a thinking exercise / commentary.

Usually, explicitly adding `bicycle=yes` should mean "there is irl signage that explicitly permits bikes", which afaik HK does not have. However, perhaps there are pathfinders/routers that defaults `highway=trunk` to be `bicycle=no`, which necessitates this unusual mapping.

Oh, and, as per note/5070064 , we still need someone to go look at how the building is like because it's covered in scaffolding, but for now I am quite hopeful ref=H is entirely unaffected.

@Chief

imo don't mark ref=H (the surviving building) as `abandoned=no`.

The situation: ref=H is evacuated, together with ref=A through ref=G. No one is inside ref=H other than perhaps gov staff.

The problem:

With the evac in place, it feels like ref=H should get `disused=no` instead of `abandoned=no`. However, because ref=H is healthy, we locally expect the ref=H evac to be temporary, which means adding `disused=no` does not seem to make sense.

A more accurate description would be ref=H is now in some sort of undetermined state, in some sort of limbo.

See if it helps understand the situation.

This is a good point. I can see `start_date` should be corresponding to "abandoned building" but not simply "building" itself.

Improved via changeset/175308323