[Talk-GB] Road Names Quarterly Project
danstowell+osm at gmail.com
Tue Feb 17 12:16:52 UTC 2015
2015-02-17 11:46 GMT+00:00 Colin Smale <colin.smale at xs4all.nl>:
> A better paradigm is that the data should be "independently verifiable from
> open sources." If the sign is wrong, it is wrong. Propagating that error
> does not change that by magical thinking.
It is true that street signs can be wrong, but other "official" data
is wrong with roughly the same frequency. There is no absolute truth
that we can appeal to. So we need a community standard for which
sources of evidence we use for OSM, and that is broadly agreed to
prefer things observable on the ground. This does not rule out the use
of common sense!
> "Ground truth" is of course no good if there is nothing on the ground - such
> as boundary lines, postcodes and even "source=local_knowledge". If there is
> no sign at all, should we remove the name from OSM, even though we, the
> local authority and Royal Mail agree that it has a certain name?
This is a straw man argument, so let's skip over it.
> This "ground truth" business needs a bit of nuance now and then. It's not
> black and white - in between there are many shades of grey, where
> common sense needs to be factored in.
Common sense, yes of course, no-one said otherwise. I used to find it
odd that OSM preferred "ground truth" over official data, but I've
increasingly come to see the wisdom of this. "Ground truth" however
does not mean purely "street signs" - it's a common-sense combination
of evidence, where we give most credit to the evidence that is freely
accessible at the location (e.g. street signs, talking to people,
looking at bus stops...). This is different from Wikipedia's
consensus, which prefers official sources rather than direct
experience - a really interesting contrast IMHO!
> On 2015-02-17 11:48, Jonathan Harley wrote:
> On 17/02/15 10:03, Colin Smale wrote:
> It's only "correct" because that's the frame of reference you have chosen in
> this case. The local authority decides what a street is officially called.
> How that is transposed to signs sometimes introduces errors, and these
> errors are sometimes volatile. The OS is not the source of the official name
> either is it?
> The frame of reference we use is "ground truth" - what is actually there in
> the physical world.
> Also, the signage at the end of the street is what visitors and delivery
> drivers see, so it's surely the most practically useful thing to have on a
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
More information about the Talk-GB