[Talk-us] Imports information on the wiki
emacsen at gmail.com
Mon Jan 2 23:50:42 GMT 2012
On Mon, Jan 2, 2012 at 6:22 PM, Frederik Ramm <frederik at remote.org> wrote:
> I think that even imports that are well executed *technically* are usually
> bad because they worsen the ratio of "mapper hours available to maintain
> data" to "amount of data requiring maintenance".
> Imports should only be allowed if there is a realistic expecation that the> presence of the imported data will lead to a growth in our community of
> about the number of people that would have been required to survey the
> imported data in the first place.
I think in this case, Martijn is reacting to the sheer number of half
completed imports and fixbots in the US that have left areas half
completed or half-right.
It would be easy to say "Manually fix the data", but I can tell you
from experience that going around and manually fixing "Rd" to Road is
not fun, and can, with the TIGER imports, be done safely (by looking
at other tags and being careful with the expansion regex). This has
been done for part of the country, but not the other part.
Similarly, here in the mid-Atlantic region, we have several imports
which have been both not-complete and done twice. I've spent many
hours manually examining two polygons of the same geometry (some which
share the same nodes, others which do not) only to remove one.
Having users do these kind of operations adds nothing but "busywork",
and is error prone (in the second case, I've removed both sets of
polygons by accident, for example).
I think Martijn's focus on cleaning up the imports, especially in the
US, should be welcome and encouraged.
More information about the Talk-us