[Talk-us] [Talk-us-newyork] Highway classification guidelines for New York State

[Jmapb]

Hybrid upstate NY and NYC mapper here.  Glad to hear this project is
still churning along. It's much changed since the last time I read it.

These comments refer to the 16:36, 7 September 2021‎ version (still
current at the moment) of the proposal
https://wiki.openstreetmap.org/wiki/Proposal:_New_York/Highway_Classification


I feel the proposal page could use more introductory material before
jumping into the weeds. It would be helpful to have a "Background"
section that describes the current state of NY highway classification
and why it's in want of overhaul, and a "Goals" section that describes
the desired features of an improved classification system, according to
the best understood consensus. I'd also like to see what previously-used
tagging standards the new proposal deprecates.

I wouldn't mind taking a shot at this verbiage, but it seems that the
bulk of this discussion has taken place on other channels so I'd
hesitate to step on toes with my out-of-the-loop feet.

This much is clear: The proposal aims to map NY State DOT's "Arterial
Classification Code" (ACC) onto OSM's highway tags (highway=* and
friends), yielding an improved representation of NY's highway system in
the OSM database, benefiting routing and map rendering.

The resulting classifications would be largely objective, with
allowances for manual tweaking in some cases. But what principles are
guiding the tweaks? I imagine there have been discussions and perhaps
consensus. I feel that these unwritten rules should be written, or at
least described.


Now, into the weeds..!

 > 1. highway=motorway

  - Anomaly #2, "single-carriageway roads that are arterial class 1",
lists several examples but does not make any explicit tagging
recommendations. Are these to be tagged highway=motorway or not?
  - Anomaly #4, "short sections of motorway that connect to trunks at
either end", seems to suggest that small islands of motorway surrounded
by trunk are correct mapping practice, for FCCs in the A10 to A18 range.
Some previous mapping standards have forbidden such islands, so best to
make this explicit.
  - Re the "ball of yarn" motorway renderings, I personally don't hate
them much. If a city is thick with motorways, that's what I expect to
see. But if a tag to filter out certain local highway=motorway ways at
low zoom is needed, suburban=yes seems like a poor choice for it, based
on the examples given. There's an opaque claim that "the analysis" (the
summary table below says "consider adding suburban=yes if ACC=2", is
that "the analysis"?) has determined which motorways should be tagged as
suburban, but the Major Deegan and BQE, for example, are in dense urban
areas. (And Flatbush Avenue Extension is not a motorway at all.)
  - Relatedly, the sample renderings linked for "Map of all motorways in
New York State" and "Map of most significant (ACC=1) motorways in New
York State" both link to the same image file, NY-motorways-all.png


 > 2. highway=trunk

  - "The status of a road being a trunk has nothing to do with its
physical characteristics and everything to do with its importance to the
road network", great to have this clearly stated; I'd suggest also
mentioning stop and speed controls just to be extra explicit, since
these might not be understood to be physical characteristics of the
roads themselves.
  - Regarding the trunk roads linking up with other highway=trunks at
state line crossings: is this a project goal? Just with highway=trunk,
or with all classifications? What about the Canadian border?
  - Regarding trunks in NYC, this merits some separate discussion and
maybe a distinct ruleset (either for NYC in particular or for dense city
centers in general.) Personally, I dislike the promotion of individual
city blocks to trunk (south of the Lincoln Tunnel for instance) and I
believe that trunk is more suitable than motorway for bridges like the
Brooklyn and Manhattan that terminate on the surface-level urban grid.
In the other direction, there are candidates for promotion to trunk,
such as the Mosholu Parkway and the Bronx and Pelham Parkway.
  - Re "There are only a handful of anomalies ... these should be easy
to patch", I'd again like to see some clarification on goals. (Based on
the examples, it appears that the idea is to avoid sub-trunk islands
with trunk on either end. The "Classification for highway=highway based
on NYS DOT codes" table under Summary references the necessity to
"connect the network.")


 > 3. highway=primary

  - Re "the network appears to be free from routing islands or
unexplained spurs", are there situations where a highway=primary spur is
acceptable?
  - The state map at this point is crowded enough that it's hard to
discern where to quibble on classifications. (The easy approach, of
course, is simply not to quibble and take things as DOT serves them up.)
  - The NYC map does indeed have a deficit of primary roads within the
boros. A few of the more prominent primaries (by current tagging) that
are missing are: Richmond Ave and Hyland Blvd in Staten Island; 4th Ave,
Flatbush, Ocean Parkway, Eastern Parkway, Atlantic Ave in Brooklyn;
Grand and Flushing Avenues connecting Brooklyn and Queens; Astoria and
Woodhaven Boulevards in Queens; Grand Concourse in the Bronx; and many
north/south avenues in Manhattan. My take here is that highway=primary
does not map well to Arterial Code 3 within NYC -- and I suspect the
same issue will arise in other large cities.
  - Re the "Primary highways in New York State" map, it appears to show
many more trunk roads (in red) than the "Map of trunk roads in New York
State" above. Are some of these red roads primary? Maybe they're the
proposed expressway=yes primaries?


 > 4. highway=secondary

  - The proposal is to tag any numbered state highway that hasn't been
handled by the above cases with highway=secondary, which doesn't sound
like a bad idea... but again, there's no easy way to compare the
proposed secondaries to the current map.
  - The summary table also adds "the four signed New York State
Reference Routes" as proposed secondaries. Which routes are these?
  - No other possible criteria for highway=secondary are mentioned. Are
there any remaining sections of US highways that might qualify? And as
with the primaries, there should at least be some allowance for tagging
highway=secondary for prominent urban roads between neighborhoods.
- For highway=secondary and below, there's no more discussion about
avoiding spurs and islands. Should spurs and islands be allowed from
secondary down?


 > 5. highway=tertiary

  - "any numbered county route not covered by any of the above rules -
including unpaved and seasonal routes. These routes can be of
considerable local importance despite the inconveniences attendant upon
their poor quality" sounds a little hand-wavy to me... yes they *can* be
of local importance, but does that translate to automatic tertiary
status for the entirety of every county road? I know a few in the
Catskills that simply dead-end into wilderness. Perhaps, as with some of
the higher classifications, the proposal should allow for discretion in
places where the objectively derived classifications don't make sense.
  - The proposal also suggests highway=tertiary for any New York State
Reference Route that hasn't merited a higher classification. But that's
it. I don't like the implied suggestion of automatic demotion for any
tertiaries that haven't met these criteria. IMO local mappers need to
have a freer hand in assigning tertiary -- in urban situations,
certainly, but also for small village main streets and such, county
number or no.


 > 6. highway=unclassified
 > 7. highway=residential

I feel the proposal gets into some murky territory by basing
highway=unclassified on commercial traffic. There has always been a
highway hierarchy with unclassified at the bottom rank, and then
residential below that, ie, not ranked at all. Unclassified is for the
most minor roads that link locations, residential is for public roads
that don't. How these classifications are mapped to reality varies
wildly over the globe (the names themselves are nearly meaningless) but
the suggestion that we should choose between these two classifications
by the *type* of traffic carried -- not by the role in linking
locations, the amount of traffic, the distance covered, or the road's
routing prominence -- seems like a sharp departure from tagging norms.


 > Expressways

Re "Grade-separated interchanges (access_control=full,
access_control=partial)", these are undocumented tags. What do they
mean? Are they part of this proposal?


 > Road Access

  - "privately owned and maintained residential roads and service ways
should be restricted to at most access=destination if they are open to
deliveries, or else the network will present problems for routing
delivery trucks", very happy to see this discussed. There are many
subtle distinctions that inform good access tagging, but in general
access=private is greatly overused.
  - "Appropriate conditional access tags should be added to ways that
are only seasonally maintained", some examples here would help
standardize seasonal tagging across the state.


 > Memorial and symbolic highway names

Both honorary_name and memorial_name are in use and seem more correct
for this purpose, so why is official_name preferable? Personally I like
honorary_name because it would cover non-memorial situations, eg, where
a living person's name is used. (Currently memorial_name has more uses,
but they're all the same road, "USS Indianapolis Memorial Highway".)


 > Routes with only a route number

This topic is a tricky one, and again, would benefit from a statement of
goals.

My own goal in this area has been to provide name tagging that will
allow Nominatim (and presumably other geocoders) to find mapped
addresses in all common forms. To this end, I'm in favor of populating
name, ref, and sometimes alt_name and short_name, on the highway ways
that comprise numbered routes. On NY 28 for example: name=State Route
28, ref=NY 28, alt_name=State Highway 28, short_name=Route 28.

The proposal suggests that the street names found on the imported
address points be used to guide the choice of name tags for the ways.
But the names found in the addr:street tags vary considerably in form --
all of the street names above are in use along NY 28 in Ulster County,
for example. (I haven't seen just the ref, eg addr:street=NY 28, on any
of dead10ck's imports, but this form does exist in hand-mapped data.)

I suggest that this proposal is a good opportunity to standardize on
accurate and useful name tags for unnamed highway members of numbered
routes. The above example might be a good template for state routes;
variations could be concocted for numbered US and county roads.

The proposal mentions that "In this situation, some mappers prefer to
enter noname=yes." This is undoubtedly true, and has some decent logic
behind it. Adding numbered name tags to the member ways is
controversial. Some influential people really hate the practice. But
this is a proposal, not a tag documentation page, so we should describe
what we want to see happen. If we want to use noname=yes, that's what
should be proposed. If we want to use name tags, that's what should be
proposed. And if we're not up for implementing unified statewide
standards for numbered route way naming, then IMO there's no use
including this section in a statewide proposal -- and we can let the
current TIGER names continue to fester.


In closing, I'm very excited for this proposal and sincere thanks to
everyone who's been working on it. Happy to help in whatever way I can.

Cheers,
Jason