<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<div class="moz-cite-prefix">On Thu, Dec 5, 2013 at 9:25 PM, Marc
Gemis <span dir="ltr"><<a href="mailto:marc.gemis@gmail.com"
target="_blank">marc.gemis@gmail.com</a>></span> wrote:<br>
<blockquote type="cite">
Bart,
<div><br>
</div>
<div>I just added a postal_code boundary for 2840 Rumst. And
yes, both the Hondstraat and Steenweg op Waarloos now get the
correct postal code: 2840. They had 2550 (from Kontich)
before. So postal_code boundaries are the solution for my
nominatim problems.</div>
<div><br>
</div>
<div>regards</div>
<span class="HOEnZb"><font color="#888888">
<div><br>
</div>
</font></span></blockquote>
On 2013-12-05 22:52, Marc Gemis wrote :<br>
</div>
<blockquote
cite="mid:CAJKJX-Q559LkOMMsUSAZQagFhUB2dpSPoKYAStkco+o7D7eEGw@mail.gmail.com"
type="cite">
<div dir="ltr">Did the same (duplicate the admin relation, change
into a postal-code relation) for Bornem and there it works as
well. Sas & Nattenhaasdonkstraat now show the correct 2880
postal code. It took several minutes though before all street
segments were updated.
<div>
<br>
</div>
<div>Since in Belgium the postal code areas coincide with
village borders, we have to double them. This 1-to-1 mapping
might not be the case in other countries. When we use those
postal code boundaries, we do not have to put the postal code
on streets or admin relations anymore. At least not for
applications that understand those boundaries.</div>
</div>
</blockquote>
<br>
I find bizarre to have to add such additional relations to villages
to get a correct postcode and to have to do it by guessing, without
a written specification explaining how to do. I'd say the proof that
it's not necessary is Dolembreux below and that if it doesn't work
in other cases the reason should be found rather than finding a
workaround and concluding that it's what has to be done.<br>
Village Boundary <a class="set_position" data-id="2792257"
data-lat="50.5375706" data-lon="5.62799963481202"
data-max-lat="50.5524444580078" data-max-lon="5.66065692901611"
data-min-lat="50.5226974487305" data-min-lon="5.59724092483521"
data-name="Dolembreux, Sprimont, Liège, French Community,
Wallonia, 4140, Belgium" data-prefix="Village Boundary"
data-type="relation"
href="http://www.openstreetmap.org/relation/2792257">Dolembreux,
Sprimont, Liège, French Community, Wallonia, 4140, Belgium</a><br>
<br>
This said, I returned to Минск (Minsk, a big city) where I once saw
things like that.<br>
<br>
<a href="http://www.openstreetmap.org/relation/59195">They of course
use boundary relations</a>, but with no subarea and a single name
on some ways (interesting to know that the borderline or Minsk is
called Minsk), <a
href="http://www.openstreetmap.org/relation/79847">they have
address type relations</a> that look a bit like the German
associatedStreet but they are different, <a
href="http://www.openstreetmap.org/relation/79847">they also have
postal_code relations</a> but look at what they contain, <a
href="http://www.openstreetmap.org/relation/295203">Автобусы г.
Минска</a> (buses of City Minsk) that seems done differently from
elsewhere and a strange route to me, etc.<br>
<br>
I compare with Moscow where I see no address nor postal_code
relations, but <a
href="http://www.openstreetmap.org/relation/85473">a strange
street relation</a>, ..<br>
<br>
No wonder that Nominatim does not work if everybody is doing it
their own way.<br>
<br>
I think OSM is going crazy. Is all that really necessary? Why
don't we first try to have it work correctly as a routing (GPS)
database? According to my tests, it is unreliable, and Guy even
added "they laugh at us".<br>
<br>
Cheers,<br>
<br>
<table>
<tbody>
<tr>
<td>André.</td>
</tr>
</tbody>
</table>
<br>
<br>
</body>
</html>