[OSM-talk-be] Postalcode 12 in Rumst - Fixed

Jo winfixit at gmail.com
Tue Jun 11 09:40:42 UTC 2013


2013/6/11 Glenn Plas <glenn at byte-consult.be>

>  On 06/11/2013 05:29 AM, Marc Gemis wrote:
>
> An update  on my problem with postalcode 12 appearing on some roads in
> Rumst when you do a seach on openstreetmap.org.
>
>  I posted my question in the help forum. One of the nominatim maintainers
> (I assume), dropped the node that was tagged with postal code 12. It was
> somewhere in the nominatim database and could not be removed by editing via
> Josm (or another editor).
>
>  Since then I found that nominatim often returns the wrong postal code
> for streets in Hingene. I assume it is caused by their "closed node"
> algorithm, which return a 9000 code, simply because it's the nearest node
> with a postal_code. I solved this by adding a postal_code tag with the
> correct number on the street itself. An associated street relation with the
> correct number does not work.
>
>
>
> Marc,
>
> Tx for the information, I'm compiling nominatim documentation about the
> inner workings and this goes against everything I learned so far,
> especially the associated street relation not taking precedence in this
> case is quite surprising to me with what I know so far.  Very
> interesting.   Now I know for sure that I don't know it all yet.
>
> By going to the nominatim code and the database schema's I was about to
> debunk my own words in the past, now I need to re-reconsider.  (I've set up
> several myself and use them in production).   Seems that noone can really
> explain in simpel steps how the inner things work so I was giving it a
> try.   Really appreciate you taking the time to share this.
>

Could it be that Nominatim is caching for too long and that it takes a long
time to take update information in OSM in account?

Jo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-be/attachments/20130611/72753d6c/attachment.htm>


More information about the Talk-be mailing list