This is a really difficult question.   One that I didn't know how to do.   Originally, I had planned on accuracy as the defining point.    Each backend would signal a master process about it's accuracy and when it got more accurate.   From this the clients would chnge over to the more accurate backends. I thought this would suffice, but the civic location idea does bring up problems.   What if different backends implement different methods?   I still vote for allowing more control from the programmers, but maybe providing examples on how to change between providers and such. Keith Preston <div>On 6/19/07, Jussi Kukkonen <<a href="mailto:jhkukkon@cc.hut.fi">jhkukkon@cc.hut.fi</a>> wrote:<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> This RFC is actually somewhat related to the earlier (hopefully even ongoing) discussion on interfaces. Selecting the backend to use is not just a matter of choosing the default one, if we want to do it right. Example: Let's say the user has the following position backends installed    1. GPS    2. Hostip Scenario 1. Application wants to know the city/country (assume civic             location is now included in the position-interface). Scenario 2: An application wants coordinates. User does not have a GPS             device with her (or she is indoors). In both scenarios the default GPS backend is unusable. Problem in scenario 1 is something that could be avoided with having separate interfaces for civic/coordinate location, but that is just a special case. Scenario 2, and many other situations like it, would require the application to call get_all_position_providers() and try another backend if the current_position()-method of the first backend failed... This is not friendly at all. Conclusions/questions: * There should be an order of preference for backends   (a list in gconf instead of just the default backend?) * Getting information from the several backends should be as seamless as   possible. I don't see why we couldn't totally abstract this in the   bindings / wrapper library?      If a method (e.g. current_position()) fails, and      if no backend was initialized with generic init-method and      if there are backends still left on the list of backends           Initialize the next backend in the list and           call the current_position method() on the new backend Is there something seriously wrong with this idea? If it works, applications could just say "I want [Coordinates|City Name|Street Address|...]" and get that information without knowing any details.   -jussi _______________________________________________ GeoClue mailing list <a href="mailto:GeoClue@lists.freedesktop.org">GeoClue@lists.freedesktop.org</a> <a href="http://lists.freedesktop.org/mailman/listinfo/geoclue"> http://lists.freedesktop.org/mailman/listinfo/geoclue</a> </blockquote></div>