The irony is that if you were to build a securities markets from scratch as an electronic system, blockchain architecture for the ledger is one way you could absolutely guarantee that the authoritative record would only be eventually consistent with microsecond trades people actually place, and brokers' own independent records would still be absolutely essential to determining who owned what when a particular acquisition was allowed to go through.
There's no benefit to trustlessness in settling a transaction when you're still reliant on a counterparty to actually make it for you, especially not in a situation like the one in the article which begins with legislation that means the regulator can determine if and when a takeover is allowed to go through. Even if it could settle transactions at market speed, there's not really much point in aiming for a distributed ledger if one party's legal status means they must be given sufficient control to hard fork it anyway.
Ethereum's raiden network can achieve market speed while maintaining the property that an exchange happens or it doesn't (a single transaction can contain the transfer of both assets being exchanged) the way it works is that transactions are made but not immediately submitted to the blockchain. If a new transaction happens it is designed so that the later transaction if submitted prevents or even undoes all previous transactions on the same channel. Combine that with funds held in escrow (on chain) and a challenge period and you have a way to transfer funds back and forth at communication speed rather than blockchain speed and if there is a dispute or the trading parties wish to go their separate ways everything becomes finalized to the chain.
In this model there would be the equivalent of brokers. These are parties who are willing to act as nodes in a network of buyers and sellers. These brokers do not take ownership of user funds though. They do have the ability to refuse to act as a relay for whoever they want to refuse but they don't have the ability to confiscate funds, do fractional accounting or prevent users from using another note to rout their transactions through.
These entities could also be anonymous because there is no need to trust them but they could also be identified and only relay transactions for KYC compliant users.
> blockchain architecture for the ledger is one way you could absolutely guarantee that the authoritative record would only be eventually consistent with microsecond trades people actually place
There are architectures that can reach consistency in seconds rather than minutes. And transactions requiring seconds rather than microseconds could potentially represent a feature, not a bug.
As for things like dividends, with a distributed ledger, the company owning the stock could issue a dividend to everyone who owns the stock at time T, and meanwhile anyone who owns fictitious borrowed shares can take their claims up with whoever claimed to buy/sell a share without entering it into the official record.
Legal restrictions against certain transactions (such as mergers or acquisitions) wouldn't necessarily prevent the ledger from serving as a purely financial record. They'd just make it more difficult to use that record for shareholder voting and other questions of control.
Bitsquare, BitShares are good examples of a decentralized exchange.
> There's no benefit to trustlessness in settling a transaction when you're still reliant on a counterparty to actually make it for you
The whole point of a blockchain is that you don't rely on the counter-party to actually make the trade for you. The exchange happens at the exact same time on a blockchain.
> if one party's legal status means they must be given sufficient control to hard fork it anyway.
I don't think hard fork is the right concept here. Hard fork implies an entire network change; I think what you are talking about is the ability for a company to issue more shares, etc... They could be given this control through smart contracts.
There's no benefit to trustlessness in settling a transaction when you're still reliant on a counterparty to actually make it for you, especially not in a situation like the one in the article which begins with legislation that means the regulator can determine if and when a takeover is allowed to go through. Even if it could settle transactions at market speed, there's not really much point in aiming for a distributed ledger if one party's legal status means they must be given sufficient control to hard fork it anyway.