Some decentralized storage proposals use block chains on top of untrusted servers in order to ensure fresh and non-forked data despite lack of trust. Existing block chains (e.g. Bitcoin) are slow and waste CPU time; Algorand aims to fix those problems; so it might be a good basis for decentralized storage.
Would a modest-sized Algorand deployment (a few thousand nodes) provide good security properties? Would the logic behind proof of stake work for an Algorand-based system that starts out small?
Wealth tends to be concentrated, so that a tiny fraction of the population is very rich. Will proof of stake work out if Algorand wealth is similarly concentrated?
Do stakes have to have real value in order for Algorand to make sense? What might cause them to have real value?
In a system with skewed stake distribution, could committees end up being mostly composed of users with very small stakes? If yes, could the good behavior that stack is intended to motivate perhaps not be as strongly motivated?
If users tend to be offline much of the time, or aren't running their Algorand software full-time, will there be problems? Could a committee have too few online members to agree?
Algorand requires "weak synchrony" for safety. What could go wrong if the network sometimes failed to provide weak synchrony? Do real networks provide reliable weak synchrony in practice?
How much work (CPU, network communication, storage) does it require for a new node to get up to speed? Do ordinary users (e.g. on laptops in cafes) need to do a lot of work to join and stay up to date?
Section 9 seems to imply that the gossip protocol is vulnerable to Sybil attacks. What is the problem? Can it be solved?
Suppose a network failure causes a small country to lose its connection to the Internet for a few hours or days. Would Algorand likely still be functional inside that country? Outside that country? What will happen when the connection is fixed?