Hi,
please don't invent new tags just to "silence JOSM warnings". watersheds are being mapped since more than 15 years without the need for such a tag.

These ways are part of a proper relation with an at least documented type. That should be perfectly sufficient to just ignore warnings of an imperfect validator.

Hi,
The type=watershed tag is obviously used for relations on drainage systems of rivers which collect all the waterway ways within the main stream's drainage basin. What I mapped is a drainage divide. It is the line of division between two adjacent basins. If there is established tagging for a drainage divide, then please provide the reference.

In fact, many of the ways you modified are part of relation (type=multilinestring to be precise) which means they formally don't need to be tagged with anything and JOSM only complains because that sort of relation hasn't been implemented yet.

It's like adding tags to a type=multipolygon relation (area equivalent) just to suppress warnings in an editor (not necessarily JOSM because that one can handle them) because it doesn't know that multipolygon members don't need any tag to be a valid on OSM.

(+1 to what Manuel wrote)

For huge features like this that partially can reuse existing objects like ridges, tags on the ways are not really helpful and all tagging should be on relations. If this is type=watershed (which is mostly used in a different sense), but e.g. here for the alpine divide: relation/14461262
Or another type like 'drainage_divide' or 'water_divide' can be discussed / propose.

I would assume that they guy who mapped the Alpine divide got it wrong. Only Americans use the watershed term for a drainage basin. For normal English speakers, it is synonymous with drainage divide. But the OSM concept of watershed is understanding it the AE way.

Regarding the resuse of existing ways, it is something that I have considered, but decided against, because I assume that it will create issues:
- Particularly where ridges are oriented normal to the major direction of the divide, the typically leaves the ridge somewhere to follow the valley divide to the next ridge. Therefore, we would need to split existing continuous ridges even if these are named, have Wikidata and Wikipedia, and Wikidata referenecs the ID of the ridge way. A pattern that we do also observe quite often is the divide running up a hill and running down again at an angle < 90°.
- Reusing ridges will significantly increase the amount of relation members
Elsewhere, relations do also often not resuse existing ways, but only the notes. For example forest and meadow relations typically do not use road ways and even if forest and meadow share a boundary line, they do nevertheless use distinct ways for their boundary lines.

Concerning the Alpine Watershed relation/14461262, they tagged the ways with type=boundary and boundary=geographical where they did not use administrative borders to approximate the drainage divide.

Is there documentation on type=drainage_divide? I was looking for it, but could not find it.