[Talk-us-massachusetts] Talk-us-massachusetts Digest, Vol 22, Issue 5

[Greg Troxel]

Angela Morley <amorley at protonmail.com> writes:

> I should have probably mentioned --- important by POI/point is easy,
> but it would add a lot of load to rendering engines for each
> address. If it's at all possible to put it in a building polygon,
> that's probably the most lightweight way to accomplish it, but like I
> said, identifying the primary residence per parcel of property isn't
> easy for a machine to determine.

A few thoughts:

  we should be clear on what OSM thinks is the right way to have
  addresses, and see if we cna do that.  My impression is that it's best
  to ahve address tags on a building, when a building and an address
  match eactly, and nodes inside the building (or entrance nodes?) when
  there are multiple.

  We need to think about the future, and whether  doing something more
  expdient makes the future harder

  about matching point/building being hard for machine, keep in mind
  that if we can write code that solves 95% of the cases, when it can be
  done reliably, and leaves others, that may be better than rejecting
  the concept of doing the matching all the time.


I will grab the state data and have a look.   I had the impression that
the address points were supposed to be at buildings because that's what
911 needs.

We may want to look at the buildings layer and see what's different from
OSM, too.  But we should be careful, because the buildings layer had
some things that don't exist that we manually deleted.  So code to
associate a point with an enclosing building, and if none, to check for
a building in the buildings layer, may be useful.