[OSM-talk] Nearby Users

moltonel 3x Combo moltonel at gmail.com
Tue Nov 19 23:05:18 UTC 2013


On 19/11/2013, Frederik Ramm <frederik at remote.org> wrote:
> Or maybe before we doctor with these details - perhaps we should first
> determine what's the purpose of the list in the first place, and from
> there then think about how to serve that purpose best?

Always good to agree on the use-case before nailing the implementation :p
Here's my attempt :

There's two type of users I want to find :
 * Users are locals or go regularly to an area, so that I can ask them to
   survey or offer to meet when I visit.
 * Users who are currently interested in an area, wether they are heavily
   contributing or not, so that they can help with general mapping or explain
   the local tagging quirks.

I'd say that many people would be placed at a different geographic location
depending on the list. They will also (what most tools seem to ignore) have
multiple locations in each list.

Another problem with the current "users near me" feature is the "me" part.
It should be trivial to replace "me" with any location.




Now that I've stated the usecase, some UI ideas :
* For the "interested users", steal some algorithms from Pascal Neis's osm
  heatmap, let those fade over time, maybe add the profile locations, and then
  reverse the search and order by who's heatmap "shines brightest".
* For the "local users" list, letting them set a location in their preferences
  still seems necessary. Just let them set multiple locations, and make it clear
  that this is about where they often are, not where they often map.
We still want
  to bias toward active people, but I think local edit history is not
a good metric
  in this case. Maybe max(last_changeset_time, last_website_login) would
  work ?

I'm not sure how a "reverse heatmap search" would be implemented. Maybe
divide the globe à la whodidit and store the top 10 users for each cell ?
That's naive never_looked_at_an_osm_db_schema me again, so feel free
to dismiss it as "can't be done". But do come back with a solution to do
it anyway 1 day later :)



More information about the talk mailing list