zerebubuth at gmail.com
Tue Dec 9 13:10:51 GMT 2008
On Tue, Dec 9, 2008 at 1:41 PM, Frederik Ramm <frederik at remote.org> wrote:
> Matt Amos wrote:
>> except when it doesn't - e.g: misspelled streets, deleted "in use"
>> streets, etc...
> A little redundancy doesn't hurt - on the contrary, it makes spotting
> mistakes easier.
and prevention is better than a cure. don't spot the mistakes - use
the API's features to help prevent their occurrence.
> And about deleted "in use" streets: If a house has a
> certain address then it has that address, even if the street which used to
> pass by the house is physically removed. There is no automatism in the real
> world that links house addresses to streets, so why should there be in OSM?
> If you remove the road next to a house node with an address, then the
> address will of course remain unchanged until this is done explicitly...
> just like in the real world.
there are two uses for addressing: navigation and geo-location. the
situation you describe would make navigation very difficult, so i
think it does not happen much in the real world.
"oh hais, my house address is so-and-so street, but that doesn't exist
any more. please go via such-and-such road."?
the two tagging methods are essentially the same (i.e: they both link
elements to a way - and not necessarily the closest way) and result,
after parsing, in essentially the same data model. the fundamental
1) the relatedStreet cannot refer to streets that do not exist and
will prevent anyone attempting (possibly accidentally) to do this.
2) the addr:street represents address information textually, and
therefore requires searching (which may fail, or may result in
unintended items found) to lookup the street.
i prefer using relations, but clearly both methods have merits. how
hard is it to support both?
More information about the talk