[Talk-us] Rewriting route direction documentation to emphasize subrelations

[Minh Nguyen]

Vào lúc 16:29 2021-07-31, stevea đã viết:
> I have yet to digest all of this and try / evaluate Minh’s tool (thanks in advance!) though I’ll weigh in that some bicycle route development (USBR 66 in California) collides with this somewhat, in that it goes against “what route=bicycle network=ncn routes do” conventions.
> 
> I find sub- and super- relations to be “my most straightforward” method of dealing with route relations, which do get hierarchical.  

What the "Route directions" article documents should already be 
compatible with what you're doing for bike routes. Superrelations aren't 
exclusive to the unidirectional route relation approach; we use them to 
split bidirectional road routes at state lines too. The article allows 
for both approaches, even if one is more robust in the long term.

Bike routes could especially benefit from unidirectional route 
relations. It's very common for a bike route to be split between two 
parallel streets. They may not even be one-way streets, so it becomes 
important to indicate both the route's overall direction and the 
forward/backward role at the same time, maybe even signed/unsigned 
status too. Unidirectional route relations long ago became commonplace 
among bus routes for the same reasons.

> The invention of cycle_network (I believe Minh’s genius) helps a great deal with this, though it has its own what might be called “messiness” (despite efforts to suggest order to the moderate chaos).

I only applied cycle_network in Ohio [1] and documented it on the wiki 
[2]; it was in use elsewhere for about two years before that. [3] I have 
no idea who first coined the tag, though NE2 used it before me in San 
Francisco. [4] It would be neat if a tool could easily find out who to 
thank or blame for coining a tag without resorting to trial and error.

cycle_network was never more than an expedient workaround for the 
limitations of network for bicycle route relations. There's some 
discussion on the wiki about whether we should try harder to fix network 
(perhaps relegating its current semantics to network:type or 
network:area). [5] I see some merit in that suggestion but would be 
reluctant to take on such a large task, given the long history of 
network=lcn/rcn/ncn/icn in the database and among data consumers. No one 
wants to be blamed for breaking OpenCycleMap overnight. :^D

[1] https://www.openstreetmap.org/changeset/13658493
[2] https://wiki.openstreetmap.org/wiki/Special:Diff/826251
[3] http://taghistory.raifer.tech/#***/cycle_network/
[4] https://www.openstreetmap.org/changeset/9563726
[5] 
https://wiki.openstreetmap.org/wiki/Talk:Key:cycle_network#What_additional_benefit_does_this_tag_provide.3F

-- 
minh at nguyen.cincinnati.oh.us