[Talk-us] Address improvement through imports?

Ian Dees ian.dees at gmail.com
Wed Nov 2 16:22:38 GMT 2011


On Wed, Nov 2, 2011 at 11:17 AM, Martijn van Exel <m at rtijn.org> wrote:

> Ian,
>
> On Wed, Nov 2, 2011 at 9:24 AM, Ian Dees <ian.dees at gmail.com> wrote:
>
> [..]
>
> > Address range information can be derived from existing TIGER data quite
> > simply.
> > However, I would argue that we should only talk about importing point
> > information for two reasons:
> > 1) address ranges get in the way of editing existing TIGER features (to
> > align a road you also have to align the two address range ways on either
> > side)
>
> I agree that for *imports* we should only be looking at point data.
> Importing ranges means having to match external data to OSM ways which
> seems like a world of pain to me. Even if TIGER ids are permanent (are
> they?) TIGER tags could have been removed, ways merged or split.. Let
> alone merging ranges from other external data where you would have to
> match by name (ouch!) or geometry (also ouch!).
>

I don't think we'd want to match address ranges with existing OSM data
since it would probably take too much time to figure out and it'd probably
be wrong most of the time anyway. But we shouldn't import ranges
(especially from TIGER)...


>
> For mapping, I'd say anything that people are willing to contribute is
> good.
>

... but yes we should take address data however people are comfortable
giving it to us if they take the time to survey it.


>
> > 2) address ranges are difficult to improve (if I wanted to map a single
> > address after a photo map trip, I would have to split the existing
> address
> > range way into constituent parts)
> > ...whereas point addresses (even if we generate them artificially from
> TIGER
> > address ranges) can easily be moved to their correct location without
> > modifying complex way geometries. Their tags can be copied on to nearby
> > buildings quickly and easily.
>
> Generating individual addresses from TIGER ranges means adding a layer
> of inaccuracy on top of a dataset that is already of sketchy quality
> to begin with, isn't it?
>
>
We're going to have to manually merge this stuff (including realigning
address points based on Bing or other aerial imagery) no matter how we
create it or where it's from, so as long as it's in the right vicinity it's
still useful to us.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-us/attachments/20111102/b7419e8f/attachment-0001.html>


More information about the Talk-us mailing list