I am still struggling to map the local bus routes where I live. It is difficult to reconcile reality with what people expect and with the OSM guidelines. The schema with which I have come up goes something like this:

map each variant of a bus route (route letter/number, direction, reverse-runs, self-crossings, short-runnings, telescoping, spoon routes, intermediate terminuses, ecs movements, letter/number ambiguity, same vehicles changing number) as a separate relation, as the guidelines say:

take, as my starting point, the block in most bus timetables where it says “.. and then at these minutes past each hour, until…”. Tag such routes with bobby444:direction=E, W, or whatever bobby444:variant=1, 2, etc

For any variations on these, outside this block, add higher variant numbers, with higher numbers corresponding to something like a ‘routing metric’, with higher numbers corresponding to less frequent, less likely, less useful, less regular variants. The variant numbers must obviously be unique, but I haven’t decided whether they should be contiguous or not.

For minor variations (the test for minor being ‘can I realistically walk from any point on this route to a suitable point on the parent route), the variants get a decimal number. Such variants might be not serving a shopping center after shopping hours, or not going into a housing estate at times when it would cause congestion. A variant can either ‘add’ or ‘take away’ ways and/or stops from a parent route.

The idea is that any simplistic route renderer will just take the routes as they are, and just overlay them so that they end up looking like a huge undifferentiated web (as now), but a more complex renderer will be able to ‘spot the difference’ and decide how to present a less popular route (dotted line, omit it altogether, show it all the time, add textual annotations, allow the user to filter it in/out based on day/time)

The namespace bobby444 won’t be suitable for long term use.