Multimap has tried using 4 various crowd source data sets and integrating it with their api.
Wikipedia data is hard to work with. As you need to import and host the data, parse the data, regulary get new dumps and re-parse and update ur data.
But the positives are that the data set size is very large. The people running Wikipedia are very nice and helpful, were not pushy about donations etc.
MM rest API is really good and provides many GIS type features to end developers. MM had done a proof of concept integration with Flickr’s web service search API. The integration does not have any mediation and as a result incorrect geo tagging goes by without checking, so incorrect lat, lons are present and nothing can be done about. As is the same for offensive photo;’s etc. Also Flickr does not allow for commercial use
Geonames provides a crowd sourced database of placenames and their co-ordinates. MM API uses the Geonames data in the background as a fallback method or a first choice method based on users input for POI searches like London Eye.
Database is constantly increasing and as a result is always uptodate and contains names that people are using. So if this part of your geocoding solution then the place names people use will come back with locations people expect.
OSM tile integration into MM JS API is another integration which was not released due to comercialisation issues. John Mckerell developed a bookmarklet which would add the OSM tile links into the MM mapviewer on multimap.com website.
Coverage is both good and bad. It is not world wide but areas that are covered are in grave detail and that is a good thing. John tends to collect all information when out mapping instead of doing a pub mapping run and then parks one and so on.
POI information in OSM would work well with MM’s API. It has intelligent spatial search methods which very neatly expose information from OSM data set. Licensing issues need to be resolved before this can happen.
Strength of crowd-sourced data is definitely around and many people are working on the same data
MM can have really positive effeects for minimal effort, benefits are ongoing as projects expand
MM can’t build sustainable diiferentiation