[Talk-us-massachusetts] more on Lynn Woods

[Greg Troxel]

Tom Parent <tomparent at gmail.com> writes:

> I do feel OSM has some glaring technical deficiencies that make it way too
> easy to screw things up...especially for casual users not using JOSM.

Yes, lots of rope is handed out :-) I more or less feel that relation
work should be done in a JOSM-class editor.  And I'm not a JOSM fanboy
-- I dislike java and I find it clunky.  But I can edit relations
without fear.   That said, it's obvious iD is vastly better at relations
than potlatch.

> OSM could greatly benefit from "layer" functionality.  IMHO landcover
> tags should be a completely separate layer where one is unable to make
> relationships with other ways (roads, cadastre, etc).  The following
> should all be separate layers: land cover, structures, boundaries,
> hydrography, transportation, other infrastructure (power lines, etc)

I see your point, but OTOH OSM has been a very interesting experiment.
The orthodoxy of GIS and layers is very well established.  I have been
working with qgis doing extreme micromapping of some conservation land
and also a couple of residential lots, and to add one of an object
(e.g. well head) the GIS way would be to define a well layer and then
put in one point.  OSM, in contrast, allows arbitrary tags on objects,
so you can add an object just by adding a tag, without defining a layer.
This is a blessing in that it makes it easier to map things not part of
the schema, and grow the schema over time.  It is a curse in that it
leads to massive semantic confusion if not done carefully -- and
landuse/landcover was not done carefully.

It seems that one could define a mapping to/from layers from OSM tags,
and that we can apply the same layer thinking in the OSM
representation.   So the layer-type thinking you suggest (and that I
agree with) could be implemented in editors.

For example:

  a layer selection box that if a type is unchecked, then those tags and
  objects that would be untagged wihtout them are hidden

  validation rules that in many cases, objects cannot have tags from
  multiple layers

But sometimes multi-layer objects are sensible:

  We add amenity tags (and I note that you did not include those) on
  buildings when that amenity is the sole occupancy.

  Sometimes, shorelines and boundaries really are definitionally
  related.

That doesn't mean a large pile of validation rules along your
suggestions won't work.

In normal GIS, cross-layer relationships are handled by snapping
options, where points that become very near in one layer are set to
exactly match the other when being added or moved.  And I think all
points moved at once when editing sometimes.  This would be the
equivalent of e.g. two ways in OSM sharing a line segment and moving a
node on one end.  It mostly comes down to the person editing
understanding the semantics and being intentional about it, and this
this pretty complicated stuff.
-------------- 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-us-massachusetts/attachments/20220728/ba557f5b/attachment.sig>