[OSM-talk] Google Maps - OSM comparison

David Earl david at frankieandshadow.com
Sun Nov 30 13:27:12 GMT 2008


On 30/11/2008 13:09, Gervase Markham wrote:
> David Earl wrote:>
>> However, if we start applying similar techniques to state captials or 
>> other hierarchies, a search inferred from a loose syntax will not be 
>> enough and I need to provide a more formal way for mechanical clients to 
>> constrain their searches. As it stands "city" is ambiguous - it is both 
>> a category and part of some names (to wit, "Mexico City").
> 
> One hint it doesn't seem to use is the current viewport. If I'm viewing
> London, I probably want places in London sorted to the top, as opposed
> to placed in Scotland or places in the USA.

There's have to be some indication that's what was wanted IMO. This 
discussion stared because people wanted (expected?) results to be 
ordered by "importance" (though we only have a very coarse metric for 
importance currently, which is why the expectation wasn't being met), 
not by proximity to location.

One reason you might be doing a search is because you want to make a big 
jump away from what you're looking at currently without all that tedious 
zooming and panning, so assuming the current view is the centre of 
interest is not necessarily a correct assumption (though it certainly 
might be sometimes)

Also, though I'm sure it is possible, and I could use optimizations of 
various kinds, ordering by great circle distance (which you'd need to do 
for this) is quite a lot more compute intensive in the ordering than the 
local linear distances I'm using to order by when you qualify by place 
("Bahnhofstrasse in Munich").

David




More information about the talk mailing list