Thank you for this explanation and for the series of excellent posts on Zilliqa technology.

The linear scaling claim seems to be based on the assumption that Type III transactions aren’t common enough to be a bottleneck. However, all decentralized exchange transactions, and more generally escrow transactions, involve 2 or more users and 1 or more contracts, i.e. fall under Type III. Is the assumption that these types of transactions aren’t common, or that they can somehow be converted to a sequence of Type II transactions in which all counterparties agree to use addresses corresponding to the shard to which the contracts are assigned?

I say contracts (not contract) because in many cases multiple contracts are used to carry out the exchange. In such cases it seems all cooperating contracts would also need to hash to the same last bits to correspond to the same shard, otherwise, the transaction would need to go to the DS committee.

It would be great to see an example of how an ERC20 token exchange transaction (e.g. 0x) would be handled by a shard on Zilliqa.