Comment by lambdaone

The same thing will likely be true of raw files, as we get better (and probably machine-learning based) algorithms for decoding those into standard image formats. I wonder what fraction of projects currently preserve their raw files?