logo separator

[mkgmap-dev] Question reg. precompiled bounds

From Thorsten Kukuk kukuk at suse.de on Wed Jan 25 18:27:17 GMT 2012

On Wed, Jan 25, WanMil wrote:

> > On Wed, Jan 25, Gerd Petermann wrote:
> >
> >>
> >> Yes, but that shouldn't matter in the LocationHook. The LocationHook just adds tags like mkgmap:postal_code to the nodes that form such a building, and I guess that persons who are interested in postal codes
> >> will use coresponding lines in their styles to analyse mkgmnap:postal_code AND e.g. addr:postcode.
> >> Right?
> >
> > I think yes ;) At least most styles have:
> >
> > mkgmap:postal_code!=*&  mkgmap:postcode=* { set mkgmap:postal_code='${mkgmap:postcode}' }
> > mkgmap:postal_code!=*&  addr:postcode=* { set mkgmap:postal_code='${addr:postcode}' }
> >
> > The only question is: should we assume mkgmap:postcode is more
> > correct than addr:postcode? I think the order should be switched
> > in the default style.
> Do you have an example where this would improve the postal code assignment?
>  From my point of view it is better to believe the postal code 
> information from the boundaries. One simple reason for this:
> A postal_code relation catches all nodes and ways and it has to be 
> tagged once. In constrast you have N elements all tagged individual.
> The chances are much higher that there are one or more typo errors in 
> the elements than there is one in the relation.

Let's say it this way: I trust the elements much more than the
postal_code boundaries in my region. There are two reasons:
- in my region nearly all postal_code boundaries are tagged with a
  fixme: boundaries are guessed and needs to be confirmed.
  I don't know if the fixme tags are still valid or not. But at
  least in the office, which is on a postal_code border, they
  don't really fit with the official addresses.
- every second week somebody breaks several postcal_codes, because
  they split/merge ways and don't care about the relations using 
  this ways.

So in my opinion address nodes are more reliable if they exist.

While I'm strictly against removing the postcal_codes from the
precompiled boundaries: Far too many addresses are missing or
incomplete, and here we can use the postal_code from the boundaries
if they are currently not broken.

But in the end: I don't care much what is in the default style,
since everybody can modify his own style how it fits best for him.


Thorsten Kukuk, Project Manager/Release Manager SLES
SUSE LINUX Products GmbH, Maxfeldstr. 5, D-90409 Nuernberg
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 16746 (AG Nürnberg)

More information about the mkgmap-dev mailing list