I agree with @cbruguera that document sharing is important to many sovrin use cases.
Most of the off-ledger storage mechanisms that you can imagine (S3, IPFS, DropBox, etc) are technically compatible with Sovrin; we’ve deliberately not ruled things out. Of course, some scenarios disclose information in a way that’s correlateable, so not every choice preserves privacy equally well. Agencies could add value here by generating a unique URL alias for access to storage, for every unique relationship that an identity owner has in the sovrin ecosystem. Agencies could also provide the storage itself (not just a URL to access the storage)–but as soon as they do that, there is a potential for a sovrin identity to lose a little bit of its portability, if content is hard to move.
In general, I think the mindset of the sovrin design with regard to storage is to use the ledger to establish secure channels of communication using asymmetric keys, and then to transmit information about documents over those channels–instead of storing the documents themselves, directly. This works the ledger less, introduces less hacking targets, and provides maximum flexibility.