[Talk-us] Special issues in LA remap

[Alan Mintz]

At 2012-06-08 15:33, Mike N wrote:
>   The original TIGER import was of poor resolution.  To set context, 
> 'blars' originally lined up the streets with aerial imagery or gps 
> tracks.  Most of his work shows up as node movements (yellow/orange 
> nodes), with a few red nodes (new nodes he added to improve geometry). 
> After the redaction bot runs, it will be clear on a grid against Bing 
> aerial that the intersections are misaligned.   To remap those ways ahead 
> of time, just center those roads and intersections against Bing 
> aerial.  Even a tiny movement will be enough to clean those nodes. After 
> a day or so, the license information should show as clean.

Is it certain that a "red" node (created by blars) can be made clean simply 
by moving it?

It seems that the easiest way to quickly clean a way that is clean itself, 
but contains a bunch of orange nodes, is to select the way and then nudge 
it very slightly (i.e. in JOSM, zoom way in so that a shift-arrow movement 
is a small fraction of a meter, but more than the resolution of the 
coordinates, which is how much?). If you are OK with the alignment of the 
road as you see it against the imagery, this would seem to be within the 
spirit of the re-licensing, and quite easier and smaller (data-wise) than 
branding everything odbl=clean or reducing the way to a stub and re-drawing 
it, which is what I've been doing for ways that are mostly orange nodes.


>An example area is here:
>http://www.openstreetmap.org/?lat=33.9687&lon=-118.208&zoom=14&layers=M
>
>    In JOSM, his edits in that area show mainly as orange nodes, with a 
> few red nodes.  JOSM's "View History" shows that most of his edits to 
> ways consisted of removing the 'tiger:reviewed=no' tag.   So the 
> redaction bot will not change those ways at all.

Really? Why does this pass the re-licensing test? I only remove 
tiger:reviewed=no when I have checked the geometry of the way against 
imagery, which would seem to be a non-trivial contribution.


>>3)  Perhaps update with TIGER 2011 data in select areas, also
>>unspecified as to exactly where.
>
>  Now that I've had a chance to look at the areas, I think I'd reserve use 
> of TIGER 2011 for only new streets or mountainous,curvy roads with 
> original poor geometry, or for streets that blars created or split and 
> show as 'red'.   For gridded streets, it's almost certainly easier to 
> just move the nodes instead of TIGER, usually to the center line of the 
> aerial imagery.

Maybe, though I'm not sure the result is better. To me, it has more to do 
with whether the area has any other user edits (other than blars). 
Glendale, in particular, may be a decent candidate for a re-import, if I 
(or someone else) can ever get around to it. I've done a little bit of 
mapping around it, but those are easy to find and merge, and a couple of 
hours to process an import sure seems better than the time it will take to 
manually go street-by-street, particularly if it is well-aligned, as it has 
been in other LA areas.

It sure would be nice if we could get some kind of commitment not to run 
redaction on the LA area until we can deal with the huge amount of tainted 
data that currently exists.

--
Alan Mintz <Alan_Mintz+OSM at Earthlink.net>