There you said it. Only on the edges of the embankment can the individual features notice that they are on some embankment.

These polygons were tagged as "warehouse buildings". This was obviously wrong because there are no buildings, but I can see the mapping intent is "these lands are for storing stuff".

Then, at minimum these are industrial landuse. Further refinement should probably stay on the basis of industrial landuse.

I view here changeset comment is the place to discuss the "small things", while Discord is the place to discuss the "big things", and when we have consensus, we post them to the wiki for long-term storage.

I feel like we are discussing the kind of question which is so obvious and trivial, it need not be mentioned in the first place.

Consider e.g. Shatin. We can very easily find photos to show that the riverside ground level is very different from the "general roads" ground level, the entire built area is basically embanked from the riverside. Then, for painfully obvious reasons and by simple intuition, all those roads and buildings need not declare `embankment=yes`.

My sentiment is not personally-directed, but this discussion is getting stupid.

I don't "still" remove embankment info. This looks pretty flat to me, there is no embankment. Past info could be a mistake.

This is notably different from e.g. Yuen Long side where I very clearly acknowledge multi-level embankment.

I will take no action. Until we have a spec for this, I will take it as "marked crossing".

From aerial imagery, this looks less like a private car park, and more like an area to store new, to-be-sold vehicles, or something similar. It cannot possibly be a parking area.

The photo quality is highly questionable, but assuming you are referring to the large sign on the left, are you sure it's not trying to describe "hard right" but actually "forward diagonal right" (which would support your gyratory opinion)?

I am heavily biased towards software solutions.

I can see there can be a case for providing intrinsic attributes (e.g. bridge=yes ) and physical attributes ( manmade=bridge ; OSM roads very often do not specify their actual width), both do not conflict with each other.

That said, look at the "5 main structures" suggested by iD: bridge, tunnel, embankment, cutting, ford. The upcoming theme is "multi-layer structures".

By definition, fords cannot be multi-layer. By practice, cuttings cannot be multi-layer, because nay deeper layer would become e.g. tunnel=yes .

We have well-known standards to describe multi-layer bridges and tunnels (using layer=* ), but so far I am not aware of any standards for describing "multi-layer embankment".

When the embankment is large enough, the features contained in the embankment need not declare that it's on an embankment. There is also this undecidability issue where the usage of embankment tag implies a certain obvious ground level, which in this case is not clear whether we are referring to the riverside ground level or the very-wide embanked ground level.

(um why are we suddenly mentioning government maps...)

I won't fully discredit the "phone GPS does not match map coordinates" observation because that's something I also observe, but I do not have conclusive finding/suggestions as to whether this is "real" and applies to entire Hong Kong, or how exactly may we obtain the real coordinates from aerial imagery.

Nothing personal, just that, procedurally, we prefer not using Bing imagery.

Hi there

It seems you used Bing imagery for OSM editing. Bing imagery is generally not recommended for OSM editing (at least for Hong Kong) because of their geographic inaccuracy. When needing to confirm geographic locations, consider cross-checking with e.g. ESRI Clarity, which is usually the most geographically accurate.

Basically, when we have explicit / separate lines to describe the direction of embankment, I think we no longer need to keep the embankment tags on the roads.

Also, it's ambiguous when the outer roads *and* the inner roads both have embankment information; the relative "layering" of the embankment is unclear. Here I just let the outer area declare that they are on embankment and just presume the inner area is on some large embanked flatland, so no need to mention "embankment=yes".

The name is inspired by way/177654101 but I can review the entire thing.

This speed is unexpected, but OK.

An implicit goal of the note is to parallelize the distorted railways, which is later done in changeset/181730371 .

Theme:

Following clarification of physical separation of Wo Yi Hop Road - Cheong Wing Road, the 18-District boundary there is calibrated to better match the road median.

Again, @HolloWorld I get what you are trying to say, but I think we have the understanding / consensus that YouTube videos should not be used for OSM purposes. While a phone's G-meter can somewhat help, I don't see how it can strongly help.

It seems to me that you are thinking that I am trying to grief the map, which motivates your "forward to DWG" opinion. While I cannot guarantee every changeset I make is consistently high-quality, it should be clear that I have been trying to improve the map by continuing to supply reasonable changesets. That's not something a griefer would do.

This need not involve the DWG (but then I do not believe the DWG is interested in this kind of small-scale quality issue).

Technically it breaks routing, but I have the suspicion that those (deleted in this changeset) paths should not be there in the first place. It is not necessary for pedestrians that want to go to Xiqu Centre to use those paths, there are obviously better options available.

These are valid concerns. I couldn't come back to this, but I later also thought to myself, even the (made by me and currently reverted) curves seemed sus.

With this reversion, I will still do another changeset to revert this reversion, but this time, I will prepare in more detail. The strategy will be to do many low-quality measurements so that the overall average cannot be too inaccurate. The previous strategy of "eyeball it once" failed because the curves here are too irregular.

@HolloWorld I get what you are trying to say, but I doubt we can use MTR Railway Protection Area charts as per copyright laws. I also did ride the trains. There may be inaccuracies, but the overall idea should be correct.

I must reiterate that the old curves must be wrong somehow because e.g. the section under Wong Tai Sin Road cannot possibly be full straight.

Note:

The exits are created without names, partly due to me not remembering them, but also partly as an opportunity to let newcomers try out OSM editing.