After looking at that YC thread I have an unhappy observation: the photo location data argument is a battle we will lose in the next two to five years if we haven't already. The right kinds of image processing will do a surprisingly good job discovering location using the photo itself based on lighting, shadows, and comparisons to publicly available images.
I'm starting to have my doubts about encryption. The protected lifespan seems to be trending shorter. Periodic re-encryption doesn't seem like a good solution for archival data, and there's the problem of getting the new decryption keys back to (possibly deceased) owners of the data...
Jonathan