[OSM-talk] Ground truth for non-physical objects

Tomas Straupis tomasstraupis at gmail.com
Thu Dec 13 10:40:09 UTC 2018


2018-12-12, tr, 15:47 Andy Townsend rašė:
> If you're looking for a project that essentially mirrors "official" data
> without actually checking that its valid then OpenStreetMap might not be
> the project for you.

  I was never for indiscriminate, automated imports without manual
checks. Accepting documents as source does not necessary mean allowing
such imports. When doing manual checks you can find (and we DO find)
errors in official documents. Then OpenStreetMap gets correct data,
not official version.

  I'm also not saying to remove the ground truth rule as such. I'm
only saying that the term "ground truth" in the context of
non-physical objects must be clarified because currently it is being
interpreted in a lot of different ways.

  What is "ground" in this term for non physical objects:
  1. Physical place which could have some traces of an actual object.
  2. Ground where non-physical objects actually live - documents.

> the general view, which I think we can see from the balance of the posts
> in this thread, is that most people back the "on the ground" principle -
> if there's a housename that looks like looks like a house name, it's a
> house name, even if it's not in an "official" list.

  Balance of posts could mean one of these:
  1. You're right and majority is against usage of documents as
sources for non-physical objects.
  2. People just do not see it possible to change your interpretation
or do not see the point in this discussion at all and simply continue
doing what they have been doing.

  But even if we would be able to vote, count, elect here in talk
mailing list, what authority would that have? In my opinion - close to
none. As in most open source/data projects people "vote" with their
actions. In this case by creating data in the OpenStreetMap database.
And most non-physical data today does not come from physical
observation.

-- 
Tomas



More information about the talk mailing list