[Talk-us] Proposed mechanical edits to harmonize county road relation network=* values nationwide

[Minh Nguyen]

Vào lúc 07:40 2022-05-02, Clay Smalley đã viết:
> There are three schemes in use for county routes, as follows:
> 
> 1. US:<state>:CR
> 2. US:<state>:CR:<county>
> 3. US:<state>:<county>

Two of these syntaxes were previously discussed on this list in 2011. It 
became clear that different syntaxes were needed because of each state's 
differing approach to managing its county routes. [1]

> The second is used in Florida and New Mexico. I'm not sure what the 
> reasoning is behind the CR infix.

The :CR: infix originated in Florida. In Florida, the counties use 
consistent signage that bears the county name. Every county assigns the 
route numbers independently; however, FDOT advocates a consistent, 
statewide numbering system. [2][3] In practice, most county roads are 
former secondary state roads that kept their numbers when transitioning 
to county control.

As I recall, mappers in Florida originally adopted Ohio's 
US:<state>:<county> scheme for their county roads, unaware that Ohio's 
situation was different. Someone advocated for a less chaotic scheme 
that could explicitly indicate a consistent county road system, so this 
infix was the compromise.

The infix allows for full-size shields to include the county name while 
making it easier for renderers to omit that detail when displaying 
generic county route shields. However, the same detail can be achieved 
anyways by moving the county name to is_in:county or relying on spatial 
queries.

For Florida, so either US:FL:<county> or US:FL:CR would make sense to 
me. Since the former secondary state routes generally cross county 
lines, the choice would affect whether these routes exist as a series of 
independent relations, one in each county, or need to be unified into 
superrelation structures.

> The third is used in every other state with county roads, including the 
> aforementioned Michigan and New Jersey. Having distinct values for each 
> county is useful in states like Ohio and Texas, where counties have a 
> lot of leeway to signpost routes however they want. But even in other 
> states like Illinois where county roads are all posted with a standard 
> blue pentagon shield, the constituent counties have some wildly 
> different numbering systems, which shouldn't all be conflated into a 
> single network=* value just because of the matching shields.

I find it especially important that the county roads in Ohio and Texas 
keep separate network values. Otherwise, if we were to give them 
identical network values and rely on is_in:county to distinguish between 
the various unrelated systems, it would be too tempting for a data 
consumer to ignore is_in:county in favor of a generic shield.

Incidentally, Phil Gold advocated for a common network value in Ashland 
County, Ohio, whose townships use consistent signage and numbering, 
about the only rational thing about how that county posts its routes. [4]

It's entirely possible that the early tagging of county routes in Ohio 
influenced mappers in more rational states to follow the same approach, 
missing an opportunity to make things more straightforward for editors 
and data consumers. With OSM Americana and OsmAnd being the only major 
known consumers of these network values so far, we have an opportunity 
to optimize the tags before more consumers show up.

> My usual criteria for whether two routes should distinct network=* 
> values are:
> * Are the routes posted with identical, or very similar, shields?
> * Are the route numbers assigned by the same government entity?
> If the answer is 'no' to either of those, they should not have the same 
> network=* value. And if both answers are 'yes', the network=* values 
> should match. Of course, there are exceptions to this sort of rule (like 
> the Trans-Canada Highway, which is tagged with one network value despite 
> route numbers being assigned by the constituent provinces).

I think the key takeaway is that network value should encode the most 
essential aspects of the route's identity. For county routes, the county 
name on the shield is sometimes essential and sometimes not. The 
<county> part of any network value should only be there to distinguish 
independent systems, not as a crutch for geocoding and not for 
expressing a mere division of responsibilities between government agencies.

[1] 
https://lists.openstreetmap.org/pipermail/talk-us/2011-April/thread.html#5671
[2] 
https://www.fdot.gov/statistics/multimodaldata/multimodal/road-naming-numbering
[3] 
https://fdotwww.blob.core.windows.net/sitefinity/docs/default-source/statistics/docs/transportation-system-jurisdiction-and-numbering-handbook.pdf#page=22
[4] https://bugs.launchpad.net/osm-shields/+bug/1025086/comments/8

-- 
minh at nguyen.cincinnati.oh.us