[Talk-us] Dirt Roads (formerly: Abandoned railway)

[stevea]

>I think the other half of the equation, however, is actually getting this
>fixed across the country. At present it appears to be just a small 
>number of mappers doing it in their areas; the US is a big place, 
>and at the current rate it's not going to be fixed any time soon. 
>Drive-by tools like
>MapRoulette are generally a good solution for systemic data quality
>problems, but in this case I think the problem's too big for that. 
>... Anything else?

I certainly agree:  the US is large, our mapper density here is low. 
That makes for slow-to-build map data.
To which Michael Patrick replies:

>Imports. The bulk of the roads in the OSM USA came from the US 
>Census, but fundamentally, the TIGER data base was primarily 
>designed to support census activities. Besides the the Census 
>Bureau, there are many other federal agencies such as the BLM, BIA, 
>DOD, etc. and their congruent state agencies that have available 
>detailed GIS dataset available. (Continues...)

Yes, this is true.  Sometimes such (federal, or state, or local, like 
a city GIS department) data are quite useful, sometimes such data (as 
TIGER) are not useful.  OSM's TIGER data, many of us agree, are 
noisy, tagged in a uniform way (residential) when that is less than 
optimal, and have been discussed many times as "need to be 
corrected."  Correcting old, noisy TIGER data is possible, even by a 
bot, but either way, manually it is huge work (and we hardly have 
THAT many willing volunteers), nor have we sustained an effort to 
take a systematic approach to entering newer, better data, which at 
least partly likely means a carefully written and deployed bot.  Yes, 
this COULD be done, and may eventually, but it is a lot of work. 
Let's consider it a medium-term goal.

In short, there are lots of good data out there that might be 
imported.  But, federal data (while sometimes good, sometimes 
bad/obsolete/noisy) often cover only federal land, state/county/local 
data are "patchy" and just that:  local, and our import process is 
detailed and takes time, people, effort, consensus and dedication. 
So, results are around what we have here in the USA:  a mishmash of 
noisy federal data that hangs over much of the rural, "TIGER desert" 
areas like an old spider web, and shiny gems of smaller pockets of 
attention (local areas, counties, even states) where dedicated 
volunteers polish up the data to be fairly useful (beautiful, 
routable, commercializable, extendable...).  Good, but we must do 
better.

There is no magic bullet.  We want excellent data that are correct 
and up-to-date.  We have a vast fifty states in which to do this: 
the fourth-largest country on Earth by area.  Yet, it has been said 
many times:  OSM in the USA has a relatively low density of users. 
Yes, our data get better, but not quickly.  Specific and targeted 
projects that identify and project-manage specific sub-areas, with 
good discussion, consensus and roll-up-our-sleeves work is what is 
going to correct this.  This will take time, let's just agree to that.

I'd like to see (more) well-identified, well-prioritized, 
even-novices-can-do-this-if-they-want such projects emerge and be 
displayed in our wiki (or someplace) so that fired-up OSM volunteers 
itching to map can "shop along the shelf," pick out a sub-project 
that gives chew-and-digest satisfaction (whether it lasts a day, a 
week or a month) and results in that warm feeling of accomplishment 
(beautiful, high quality data as useful results) once done.  Now, 
THAT'S a crowd-sourced mapping project!  We're getting there, though 
in a low gear.  Discussions like these, some identification, some 
organization, some inspiration, and we will rev it up faster. 
Elephants are best eaten one bite at a time.  (A metaphor, not 
literal)

SteveA
California
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20140902/62d35aa2/attachment.html>