Last modified: 2014-10-28 08:34:19 UTC
In order to improve visualization, according to the new UI, each entity is supposed to feature an image just like it features label(s), description(s) and aliases. As there may be cultural differences, it should be possible to specify an image per language and language fall-backs should be resolved. The data type images are stored with should probably be a setting--currently, and in the scope of Wikidata, only "Commons Media" would be a proper type, but for other use cases, some "Image Link" or "Image File" type would probably be more appropriate. It would be best, to store the image(s) using the same structure like label(s) and descrition(s) adding the image(s) to the "fingerprint".
As discussed I think we should be using existing statements and harvest them for the image to show in the header. Adding an additional field introduces a lot of work for the community for very little benefit.
Thanks for pointing out one of the worst possible solutions again. :-/ Still wondering how users are supposed to understand that mechanism.