Only connects to expressway Fanling Highway.

Why did you remove `bridge=yes`? Just add `layer=`.

Imo that's still a car park. See e.g. various franchised bus long-term parking areas (not depots).

I don't understand. It's already documented that they aren't `crossing:markings=`. Is that not a spec? What else do you want? But, ok, I guess we can discuss on Discord first; I see a thread has been created.

I went there recently, and I can tell you that this is an embankment.

I will just restore `embankment=` with `source=survey`.

Hmm, aerial imagery is not very clear, but some `natural=wood` and `landuse=grass` looks like they should be `natural=tree_row` and `natural=shrubbery` instead.

I will not comment on the obvious issues of drawing separate ways without physical separation and curved turning paths.

Not super familiar with this stuff (maybe @Kovoschiz will have something to say), but iirc look-left look-right markings don't count as `crossing:markings=`.

Imo just use `amenity=parking` + `access=private` for car storage.

(I mentioned out-of-copyright maps as an easy way to get correct geometry, not sure how @vectorial8192 couldn't comprehend that)

Relative accuracy should be more important than absolute. Even if there is really some distortion in osm data, as long as they are relatively uniform and small, it should not make a difference.

But, specifically re @vbertola's Cheung Chau example, it's a hilly area, so you should not expect any imagery to be accurate.

This should be `informal=yes`, right?

Esri (not Esri Clarity) should be clearer and more accurate than Bing.

A few meters is well within the normal error of a smartphone GPS. The general alignment issue is very likely to be an issue (or just hardware limitation) with your phone. Try walking a few hundred meters and see how noisy phone GPS paths are.

Indeed.

Additionally, for areas as old as this, out-of-copyright maps are also a great option for cross-checking. It is not common for other data sources to perform better than government maps.

> I think we no longer need to keep the embankment tags on the roads

That's just not true. `embankment=yes` exists to specifies that the section of road is on an embankment. You can't expect data consumers to look up all instances of `man_made=embankment` and guess which `highway=` is on it. You wouldn't remove `bridge=yes` or `tunnel=yes` after drawing `man_made=bridge` or `man_made=tunnel`, would you?

Re second paragraph, "on some large embanked flatland" should logically imply `embankment=yes`; I don't see why "no need to mention".

Imo the `name=` is descriptive. Also, this should probably be `historic=milestone`.

P.S. 8 mi is from Star Ferry via CPR; 48.5 mi is from Star Ferry via Tai Po Road and Sheung Shui.

I am with @HolloWorld on this one.

------------

Re Railway Protection Area, Kovoschiz once DM'ed me on Discord: "For railway protection plans, the copyright of maps in law is a more peculiar question that I can't confirm much. While, "according to government information" is usually can't be used, and other maps or unknown copyright are assumed can't be used. Also funnily, law text itself is somehow copyrighted for reproduction or publishing, but it should be uncopyrightable and usable."

------------

But still, regardless of their copyright status, this changeset is still questionable:

> if segment at Chuk Yuen Estate is full straight, then why no extra station? therefore, not straight.

What? Just because a section with a station would be straight, does not necessarily mean a section without a station would not be straight. That's fallacy of denying the antecedent. I thought you said you often do logic. If you really want to know why: https://zh.wikipedia.org/zh-hk/%E9%A6%AC%E4%BB%94%E5%9D%91%E9%81%8A%E6%A8%82%E5%A0%B4?useskin=vector#%E7%88%AD%E8%AD%B0

> railways do avoid e.g. Hsin Kuang Centre, but the fit is still too tight.

What about Upper Wong Tai Sin Estate? As mentioned on Discord, the Married Quarters is on 20 mPD while Upper WTS Estate is on 23 mPD; existing osm data says the Married Quarters are 25 storeys while UWTSE is 42 storeys, so UWTSE actually has a higher likelihood of intersecting with the tunnels.

> by survey, the left/right turns in opposite directions do not match each other; therefore, obviously asymmetric.

The number of left and right turns in this changeset might be correct (I successfully verified that), but the radius might not be. (see further below)

------------

So basically there's no evidence that this changeset is a net improvement compared to the old data. Then, simply do nothing and assuming the existing data is correct. If you still make a questionable change anyway, imo it deserves to be reverted. I only didn't do it myself because I don't have any better data, and wanted to avoid an edit war/community schism. I welcome @HolloWorld's reversion.

------------

Some additional concerns:

- According to wikipedia, an SP1900 car is 24-25 metres long, but @vectorial8192 thought it was 22.

- @vectorial8192's measurement method relies on eyeballing, which has low resolution and accuracy.

- The measurement method has not been tested and verified to be accurate/useful beforehand.

It was decided to change it changeset/181356680

I have pinged you on Discord for more info.

Imo this is a proper mainline of Lung Kui Road.

I do know that `note=` is internal, but as I am completely unsure how to tag this, I just left a `note=` for future mappers to figure out.

Hi there, could you please not use `local_ref` for refs that area clearly not local to anywhere? You can use `ref` instead. Thank you.

As said, NSI simply collects the most commonly used tags, which can be contradictory or incorrect or include typoes. E.g. `network=九巴龍運 KMB & LW` is also documented in NSI. If you don't have an opinion on networks then I would simply suggest you don't change it.