I find it incredibly telling that not a single person opposed to separate sidewalk mapping to any degree replied with a substantive answer to this post of mine which disagrees with the obvious answer.
The thread is locked now, so I will repost the contents:
Let’s say we have a situation on-the-ground like this:
To start, we only have the houses mapped:
With this data, a router can only make routes “as the crow flies” - not perfect, but functional and reasonably accurate:
Now, let’s say someone comes in and adds only these roads, which are in fact present on-the-ground, but the mapping is incomplete:
Our router’s output, now, is technically valid (it’s doing the best it can) but produces worse results than the no-street-mapping case we saw earlier:
Of course, were the roads fully mapped, we’d get the most-correct route:
So, the question is: Should the snake-like road mapping be deleted because its accurate-but-incomplete mapping results in worse routing?
My answer to that prompt is “No.”
In line with that, I object to the removal of as-termed “decorative” sidewalks.
Note on non-sidewalks
This conclusion does not apply to “non-sidewalks” - the very small (essentially an extended curb top surface) areas shown in some images posted earlier are not functionally or structurally sidewalks, and should not be tagged as such. Edge cases of barely-usable sidewalks are better solved by tagging details like width. Sidewalks that do exist and are permanently unusable should be tagged in a way that indicates that status, such as with a lifecycle prefix.
If you repeat the “It (can) make routing worse!” argument without substantively addressing this post, it is clear to me that you are arguing in bad faith.