[OSM-talk] TIGER cleannup

[Paul Fox]

beej wrote:
 > Paul Fox wrote:
 > > but what about areas where the TIGER data is simply mis-aligned?  for
 > > instance, currently the TIGET data for san mateo county is off by about
 > > 200'.  and i think (i don't live in that area, so i'm not sure) that it's
 > > the entire county that's affected.
 > 
 > I'd be surprised if it's the entire county uniformly, since that's not 
 > the experience I've had in Alameda Co (different alignments in different 
 > places).

you're probably right.  i've observed misalignment in some places, but
have never checked a whole county.  too bad -- it'd probably be easier
to fix if it was the whole thing.

 > 
 > I remember reading on the census website somewhere that the map data 
 > came from a variety of sources, and no particular datum was used.

i've actually heard that since it was delegated to the counties,
it was often entered by summer interns, who traced existing maps. 
given the inaccuracies, i'm inclined to believe it.  in the east
the problem isn't misalignment, but the fact that it seems the tiger
maps show every old farm-trail or track that's ever existed in
the last 400 years, even if it's been impassable and unmarked and
on private property for 150 of those years.  :-)

 > > i think OSM needs to have a plan for dealing with updates for these
 > > sorts of issues, at least.
 > 
 > Some options that come to mind:
 > 
 > 1) Don't ever update.  Let the OSMers do the updates.  Even the new 
 > census data at Lava Beds National Monument has errors in it (incorrect 
 > road names, foot trails marked as residential, old non-existent trails 
 > in data, missing trails, etc.)  In Chico, the new super-accurate data 
 > doesn't render or record freeway interchanges correctly.
 > 
 > 2) Update everything that's not marked as "reviewed".  This seems like 
 > it would have hellish merge implications that I'm not sure are readily 
 > surmountable.  Plus I forget to mark stuff as reviewed half the time anyway.
 > 
 > 3) Interactive regional updates... that is, maybe you get a chunk of new 
 > data, you download the OSM data, you run the as-yet-unwritten Magical 
 > Munger Program, and load the result back into JOSM.  Then you massage 
 > the data, verify it, and upload it.
 > 
 > I kinda like #1.  I think with more OSMers coming online all the time, 
 > it should be possible, too.

you might be right.  it's certainly the path of least resistance. :-)
i'm very new to OSM, so i don't have a feel for what can and
can't be done readily.  i was hoping that somehow changes to
certain kinds of imports could be marked, and somehow "reapplied"
to new data.  that all assumes that lines/streets/etc all
maintain their identifiers from one import to the next.  but i
understand it's complicated, and perhaps not worth it.

paul
=---------------------
 paul fox, pgf at foxharp.boston.ma.us (arlington, ma, where it's 30.6 degrees)