[Talk-us] Fwd: Feature Proposal - RFC - Directional Prefix & Suffix Indication

[David ``Smith'']

On Mon, Aug 23, 2010 at 3:59 PM, Mike Thompson <miketho16 at gmail.com> wrote:
> You have caused me to do some real thinking and digging on this
> subject.  I think this debate is good, and will lead to better OSM
> data in the long run.
>
> If official local government sources use pre-directionals, this is
> very strong indication that they should be part of the name.  The
> Franklin County Ohio (where Columbus is located) Highway engineer
> includes directions in some cases on their maps. Note "E. North
> Broadway" and "W. North Broadway" on the following map:
> http://www.fceo.co.franklin.oh.us/Map-Atlas%20Pages/map_19.PDF, but
> not all.  Note "W Broad St" in this map from the same source:
> http://www.fceo.co.franklin.oh.us/Map-Atlas%20Pages/map_27.PDF. Also,
> the Franklin County Auditor often uses directionals in their property
> database, although not consistently.

Holy crap, that map is available online???  I've grown up with that
map, and I've probably seen just about every grid square of every
edition since the early 90's, but I never knew it was online!
Anyway... I can definitely offer some relevant information.  First of
all, that map isn't actually produced by FCEO.  The paper editions
have for several years said "Copyright Seeger Map Company".  The data
comes primarily from Franklin County Auditor's GIS, I believe.  Seeger
supplies the exact same map to AAA with different colors (the most
significant differences being the line colors of freeways, and swapped
background colors for Columbus and unincorporated Franklin County).
Seeger is not local; they are based in Wisconsin, and I have seen
their name on a AAA map of Sevier County, TN which was not up to the
quality I've come to expect from their Franklin County maps.

Now let's take a look at that map.  That page 19 is a good place to
start.  Yes, North Broadway appears with directional prefixes.  So
does Woodruff Ave, at least on its W half.  But none of the other
east-west streets on that page are shown with directional-prefixes.
As you may have guessed from TIGER, *every* east-west street that
crosses High St actually has a directional prefix.  Seeger omits most
of them, so that could be an argument in favor of leaving them out of
OSM street names; on the other hand, Seeger isn't entirely consistent
about it, so that could be an argument that the map isn't credible
evidence for either case.

Now let's move on to matters of signage.  The City of Columbus is
extremely consistent with its street signs.  They are all-uppercase
FHWA Series C or D; occasionally Series B might be used for very long
names.  (I know of a single instance of mixed-case Clearview, and
several in mixed-case Helvetica that were apparently part of a
multi-jurisdictional experiment, but I digress...)  Street signs in
Columbus are always white on green, except a few on OSU campus that
are white on red.  Columbus tends to display one street name on each
corner of a 4-way intersection, so the name of the cross-street can be
seen on the near-right or far-left corner.  Information displayed on a
City of Columbus street sign is as follows: [dir. prefix*] [core name]
[type suffix*] [dir. suffix*].  (*Where applicable: type suffix almost
always exists; dir. prefix common; dir. suffix infrequent.)  (I've
actually wondered why Columbus doesn't put address numbers on the
street signs...)  The core name is the only part displayed at
"full-size", though it's not "too big" like on the street signs in
Westerville.  Everything else is about half the font size as the core
name, and all the text is aligned to a common midline.  Suburban
jurisdictions tend to apply variations: some use mixed-case text, some
apply the same font size for the whole sign, many employ alternative
colors, and a few use non-standard fonts and "icons"; still, most are
consistent across their entire territory.  Franklin County puts the
county (or state or township) route designation in tiny text below the
name, similar to most county signage practices in central Ohio.
Delaware County seems to have guidelines that even apply in
municipalities, apparently favoring backlit mixed-case Series E(M)
hung from the same mast-arms as traffic lights, for major
intersections.

Finally, local usage.  It's common to hear about South High Street or
West Broad Street, particularly when referring to places far from
Downtown.  This may be a matter of collective habit; since those
streets are so long, perhaps people tend to think of each of them as
two or three completely distinct streets.  (Based on typical casual
usage, High Street seems to become North High Street somewhere north
of OSU, while it apparently becomes South High Street as soon as it
crosses I-70.  Broad Street probably becomes East Broad Street at or
near the Alum Creek Bridge, and it apparently becomes West Broad
Street in the Hilltop neighborhood.)  North Broadway is another one
that frequently has its directional prefix attached, though it makes
for an awkward concatenation of directions.  I can't think of any
others where the directional prefix is commonly used.  However, I can
think of several where use of the directional prefix just sounds
strange: James Road, Lane Avenue, Dublin-Granville Road, Fourth
Street, Fifth Avenue, Eleventh Avenue, Seventeenth Avenue, Wilson
Road.  I think the roads with prefixes attached are often the ones
that have a lot of business on them which advertise their addresses in
the media; conversely, roads that are mostly known as a means to get
somewhere, rather than a destination, are better known without their
directional prefix.  As for residential streets with directional
prefixes, I don't have first-hand knowledge of how often they're
actually used.  Most of the residential streets I'm familiar with
don't cross Broad or High and are therefore prefix-free.

Now it's time for my opinions regarding OpenStreetMap.  I'm mostly in
agreement with Kevin, but that's more a matter of pragmatism than
passion.  Before I conceded the abbreviation argument, I was perfectly
happy with names like E 11th Ave.  The only time I would change a
TIGER name was to replace Pky with Pkwy.  But I realized I was in the
minority on that issue (either that, or those against abbreviations
were just being extra assertive) and that's when I started putting the
abbreviated name in abbr_name=*.  This also meant that the name=*
value wouldn't necessarily match addr:street=* when address data is
added.  When I saw someone else make a passing comment (that may well
have been Kevin) about directional prefixes only being useful for
addresses -- an opinion I already held -- I decided they could be left
out of name=* in central Ohio.  As long as the eventual addr:street=*
still included them, there shouldn't be a problem.  This newfound
freedom to express my local knowledge along with the task of expanding
abbreviations is what led to that previously-cited changeset in
downtown Columbus.  I also left the directional prefixes out of
abbr_name=*, mostly because had I left them in, abbr_name=* would
often be almost as long as name=*.  I definitely didn't want to expand
directional prefix abbreviations for name=* in most cases, because
that would cause confusion regarding roads that have directional words
in their cores, like West Case Road, North Star Road, Westpark Street,
North Broadway, etc.  Still, I'm not entirely satisfied; I started
using abbr_name=* to satisfy the concept of storing "exactly what's on
the sign", but that's not quite achieved if the directional prefix is
left out.  Maybe we should use signed_name=* or name:signed=* to store
exactly what's on the sign, preserving abbreviation and prefixes where
present?

Between that and the proposals to break the name into different tags,
what happens to good old name=*?  Does it hold the full unabbreviated
name, just the core of the name, or something in between?  Pragmatism
suggests that name=* should hold a "default" name that makes sense
with no processing or parsing, to accommodate applications that don't
understand more complicated tags.  (As far as I'm aware, at the moment
that's nearly all of them.)  This is also the reasoning behind my
stance on ref=*, and any other tag that may be expected by some
editors to convey complicated or structured information.

Now I'd like to address that argument about font sizes, prefixes, and
driver confusion.  Let's say a driver is looking for North Wilson
Road.  The sign actually says N WILSON RD, with WILSON twice as large
as the rest.  If their OSM-generated directions say North Wilson Road,
they may or may not recognize quickly that the name on the sign is the
same as the name on their printout.  They don't look very much alike
to me.  Right now, OSM has the TIGER-endowed name=N Wilson Rd.  That
looks a lot like what's on the sign, and I think it would lead to
successful navigation.  It's my understanding that there's a bot
running that would change it to North Wilson Road.  If I get to it
first, and haven't changed my mind on the matter, then it will get
name=Wilson Road.  That's fairly similar to what's on the sign, and
exactly what locals call it.  So if the driver has to stop at a gas
station and ask for directions, asking for Wilson Road will result in
the least confusion.  If their OSM-based routing application actually
looks at more than just name=* -- a highly unlikely scenario -- then
the printed directions might have the road name formatted any number
of different ways, with varying potential usability.

More pragmatism ahead: A map's main purpose is to help someone get
from one place to another.  Changing Rd to Road when the sign says RD
is slightly unhelpful for this purpose.  Removing the N from the
beginning, when it's present on the sign, is also slightly unhelpful.
On the other hand, changing the N to North is a bit more unhelpful
than omitting it, because it looks more like a word that's part of the
road name.  For nearly all uses of directional prefixes in central
Ohio, that's not the case.

PS: I'm seriously considering doing some major work with addresses
around Columbus.  As long as it's taking to make a satisfactory TIGER
address import bot, I might as well just sit down with Potlatch and
ArcExplorer's Identify tool to manually add address interpolation ways
with help from Yahoo! imagery and my familiarity with the Franklin
County address grid.  I can also conduct in-person surveys and
verification along key roads.

-- 
David "Smith"
a.k.a. Vid the Kid
a.k.a. Bír'd'in

Does this font make me look fat?