So, you suggest that we follow that parallel, and that sidewalk=yes
should (if one cares about extra detail) be mapped as sidewalk=track
or sidewalk=lane
(the same as cycleway=yes
is in more detail mapped as cycleway=track
or cycleway=lane
)? Or am I misunderstanding the meaning here (it seemed so far to me that you oppose that similarity between cycleway and sidewalk)?

If we consider every part of a street where pedestrians can walk on, a sidewalk, then I think some things end up with weird definitions. And then “sidewalk” should be changes to “footway”, so that would be more clear
would it be more clear? According to wikipedia (and industry standards in Great Britain that someone linked here earlier) “sidewalk” in North American English means exactly the same thing as “footway” is in other English dialects – " path along the side of a street, highway, terminals."
So, it seems common meaning of English word “footway” is different than what OSM means by highway=footway
(which encompasses both Sidewalk and Footpath).

(e.g.
footway=right
to indicate that pedestrians walk on the right side of the road).
Also, according to its wiki “Key footway=*
refines the tag highway=footway”
So even if were wanted, it could only be used for separately mapped highway=footway
, not to pedestrian way mapped as an property of highway=residential
. But as you note that you don’t actually want that; I agree. It is long standing tradition in OSM that separate lanes
(i.e. parts of the carriageway with separate semantics but separated only by paint) are not to be mapped as separate OSM ways.

Anyways, to me, the dialectal vocabulary and wiki history are secondary.
I completely agree with that!

The most important consideration is whether the tag’s definition is practically useful
I agree too it is very important. However most important consideration (to me) is whether wiki definition is consistent with how tag is being used in practice. If theory (wiki) and practice (OSM tags in DB) diverge; the options forward are:
- document that the tag has multiple meanings. Easiest, but reduces the usefulness of the tag for all data consumers (but that is not because of documenting multiple meanings, but because of original wording led to such ambiguous use!)
- invent new tags that are not ambiguous, and deprecate old tag for all new uses (preferably “enforcing” it via editors warning), and slowly replacing old tags until it is gone completely (improves the usefulness, but takes a long time and much effort)
- clear up the confusion in wiki, and manually verify (by contacting original creators, or resurvey) ALL existing usages (obviously, only after wide community discussion – this is mostly doable only in cases where tag usage is quite small, so probably not an option here)

If defining
footway=sidewalk
in terms of function causes this pedestrian lane to require a bunch of tags to undo the essence of whatfootway=sidewalk
was approved for – physical separation – then this is classic troll-tagging
I’m afraid I lost you there. Can you give concrete full examples what you meant here? I.e. which set of tags would you find OK, and which set of tags would you find as trolltag (and why)?

For the pedestrian lane, I favor
footway=lane
on the roadway (not a separate way), because it follows existing patterns and is unlikely to trip up any data consumer that has been interpreting OSM tags correctly.
Isn’t footway=*
intended to be used exclusively as property of highway=footway
? Both its wiki (which I linked above) as well as common OSM practices (e.g. if main tag is foo=bar
, then bar=*
property tag only refines that foo=bar
, and is not to be used on foo=xyz
). Or am I misunderstanding how you would use it?
So, IMO, adding footway=something
on e.g. highway=residential
would definitely be breaking existing patterns and not following them (it would also break explicit wording of Key:footway wiki). Such cure might be worse than a poison.
footway=*
should only be used on highway=footway
(but then, in the specific case of footway=lane
, we get into much bigger trouble of mapping lanes as separate ways – which I think there is strong consensus we don’t do in OSM. And also would be impossible in case where pedestrian lane is in the middle of residential road)