[Talk-us] Fwd: [OSM-dev] Licence redaction ready to begin

Kevin Kenny kkenny2 at nycap.rr.com
Wed Jul 11 15:20:21 BST 2012


On 07/11/2012 09:31 AM, Frederik Ramm wrote:
> Obviously my comment was a bit tongue-in-cheek; I am personally
> convinced that the unedited TIGER landscape - i.e. a map of which
> virtually nothing is correct and once you start to work somewhere you
> have to touch almost every single object if you want an acceptable
> result - is the worst situation for attracting mappers. Therefore,
> returning an area to how it was after TIGER (and deleting selected
> objects for good measure) is certainly not creating an "empty area" in
> the sense of the "do empty areas attract contributors" theory.

I'd be fine with trying to clean up unedited TIGER - or even a
"horrible mess of half-deleted TIGER" - if the expectations of
the results weren't quite so daunting. We seem to have come
around to seeking perfection rather than improvement - which
is death to crowdsourcing. The culture of
OSM seems to be veering from "bad data happens, and when it does,
other mappers fix it," to "we have to protect the map from the
mappers."

The data checks in JOSM and Potlatch2 are fine in that they all
indeed highlight potential problems. But the sum total of them
is just overwhelming. Right now, it feels as if I need to rectify
every problem in any object that I've downloaded, to silence
all the complaints from the tools. They seem to lack even the
idea of "forward motion" - this part of the map isn't perfect,
but at least it's better than what was there before. Instead,
the UI at default settings appears (to a novice) to insist on
perfection.

This becomes even worse if someone is working on, say, a piece
of the NHD import. There seems to be the expectation that if
I'm importing a sub-basin, I'll go out and visit every place
where a road and a stream meet, to get the levels right and
either mark the road as a bridge or the stream as a culvert.
If I fail to do so (and that fine distinction may not be
obvious even in the field!) I'll get raft of warnings for crossing
ways. I won't have broken rendering. I won't have broken routing.
But I'll have failed to verify personally that the resulting
rendering and routing are actually correct in every detail.

The fact that TIGER was of questionable quality and was done as a
bulk import isn't really the problem, as I see it. Even half-deleted
TIGER isn't really the issue. We'd have had
exactly the same situation if the street network had been created
by an army of novice mappers doing their own neighbourhoods - and
possibly leaving part way through the job: a
mountain of data of variable quality and completeness, and a
scolding from the tools
when you attempt a small fix without cleaning up everything.

How do other mappers deal with the perfectionism that pervades the
community?  I'm actually starting to keep tables in my personal
PostGIS of "things that I've mapped, but can't push upstream
without fixing tons of things that I haven't mapped."  Or
"things that I've copied and pasted from public domain sources,
but haven't verified with boots on the ground." Or "things that
I can't quite figure out how to tag."

I know I've been advised in the past, "just do it. People complain,
but they don't revert good data."  Is that still sound advice?



More information about the Talk-us mailing list