[OSM-talk] Good practice, and should we rely on defaults?

[Greg Troxel]

Martin Koppenhoefer <dieterdreist at gmail.com> writes:

> There are other defaults that are contested, for example lit on roads. Some
> people think the default is yes (for example for residential roads), and if
> you believe it is like this, you have to tag lit=no on the others, or it is
> "no" and you have to tag lit=yes.

This is an interesting example.  The idea that lit=yes is a default on
roads seems crazy to me.  It may be true in the city, but where i live
the vast majority of roads are not lit -- just some but not all
intersections.

> There is also the StreetComplete idea. The reasoning is, either you have
> the information, or you don't and have to guess what is likely, but
> basically have no clue whether it is the "default" or is missing the
> information. Therefor they encourage people to add a lot of explicit
> information from survey, that could have been guessed with a high
> probability but still some uncertainty.

There is some merit in that.

My view on defaults is;

  1) First, as Frederik says we need a machine-readable, group-maintained,
  consensus definition of defaults, presumably organized by some admin
  boundary (or perhaps other polygons? maybe not).  Renderers including
  routers should be encouraged to use this file, and treated as buggy if
  they do not.

    (a) Doing this will expose that it much harder than it seems.  For
    example, in the US one might expect footway to be paved and path not
    to be paved.  Unless one lives in the city in which case both are
    expected to be paved.

    (b) This file should differentiate between "if not explicit, assume
    X" and "and further, not X is extremely rare *at all locations for
    which this rule applies*".

  2) Until we have (1), both published and in significant use, it is not
  reasonable to tell someone that is adding a (true, observed0 default
  value to an object that they are doing something wrong.

  3) Even with (1), explicit tagging is often useful.  If the default is
  overwhelmingly true, then not tagging is reasonable.  If it's not
  >=99% in any given region, it seems like bad practice to encourage it
  not to eb listed.

  4) A practice of not tagging defaults is more reasonable for things
  that are so overwhelmingly important that it's close to unthinkable
  that a non-default value would not be tagged.  A good example is
  oneway=no on roads.   We as a group are ok with "missing oneway is a
  bug if oneway=yes is right, and it's better to just insist on fixing
  that rather than adding oneway=no on roads known to be oneway".

  5) Defaults are often coarser than what we tag explicitly.   There's
  surface=paved which might be a default for roads in many places
  (e.g. US, EU).  But there is surface=asphalt and others, which have
  more infomration.

  6) OSM is a collection of a vast amount of information.  Many people
  find some of that information silly to record.  But with all of us
  entering what we care about, there is a vast amount of data available.
  Tolerating data that a person doesn't care about (that conforms to OSM
  norms) is the price of both a healthy community and the benefit of
  data that others add that one does find uesful.  So I am very
  reluctant to say "don't add tags for that" in many cases.  But "don't
  tag access=yes on highway=primary" seems fair; see 4.
  
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20220406/7f45af4c/attachment.sig>