[Imports] Large scale building edits in Fiwila, Zambia

[Frederik Ramm]

Bjoern,

On 06/11/2017 11:17 PM, Bjoern Hassler wrote:
>     I think it would be better to trace actual building outlines rather
>     than generate generic placeholders from single nodes, 
> 
> Of course. However, it was a significant effort to mark the 3200
> buildings, and it's impossible for me to manually trace 3200 buildings.

Then trace as many as you can, and leave the rest to people who actually
have the time to do it right.

With JOSM's building plugin I can trace a (square) building with three
mouse clicks. If the imagery is good I can easily trace 10 buildings per
minute. I need less than a working day for 3200 buildings.

What you propose here sounds awfully complicated and will very likely
involve more than a person-day of work.

Adding extruded buildings - i.e. a pure guesswork building outline based
on a centre point someone has manually placed - is adding fake
information to OSM. Don't do it. If you only have the time to place a
centre point then that's ok, tak the centre point as building=yes, but
don't add a builing outline that is guesswork.

Did I understand you correctly that you even have developed a plugin to
extrude buildings? And at the same time you're telling me you cannot add
3200 buildings by hand? This only makes sense if your plan is to, down
the line, repeat this procedure in even greater numbers.

> Only from a certain perspective :) The health workers we work with need
> to primarily know where buildings are.

Then the points are sufficient.

> Of course, the nodes would
> suffice for that. However, node[building] doesn't render in standard
> carto, hum carto, or maps.me <http://maps.me>. 

This is not an acceptable reason. You are planning to add fake data just
because you (a) cannot be bothered to spend one working day to add
correct data, and (b) you cannot be bothered to set up a map rendering
that works for you with the data you *can* add. You are trying to abuse
our database for your project.

> So what's the practical solution, for us to support the health workers?

Either map real buildings correctly; or map buildings as points (and if
you need it, set up a rendering server that displays little house icons
where these points are), or use something other than OSM.

Setting up a rendering with "house icons" for an area as small as the
one we're talking about should be easy with a tool like Maperitive; you
don't even need a full-blown tile server.

Don't add the "olc:short" codes either; as you have been told, anything
that can be computed from the lat/lon unnecessary.

If you add the data in the way you have described here, it will most
likely be removed again. It would be reckless to build a workflow your
health workers depend on that includes abusing OpenStreetMap, because
the workflow *will* break - and that will then not be the fault of the
OSM mapper who has removed your buggy data, but your fault for assuming
you could add buggy data to OSM!

Bye
Frederik

-- 
Frederik Ramm  ##  eMail frederik at remote.org  ##  N49°00'09" E008°23'33"