Let's take take a textbook example of HFT with orders larger than single exchange can satisfy therefore overflowing to other exchanges and examine timeline. 1. Investor places order at exchange A --> 2. order is known to both exchange A and HFT --> 3. HFT places reactive order in exchange B --> 4. exchange B receives "overflown" order from exchange A.
The time between exchange A receiving original order (point 2.) and exchange B receiving remains of that order (point 4.) is extremely small, but larger than zero. The argument here is that until exchange B acknowledges reception of the original order in its ledger, that information is not public at exchange B and acting upon that order in exchange B until exchange B acknowledges said order is acting on non-public information, therefore front-running.
> <...> the order flow is technically “public” once broadcasted by the exchange.
The core argument I see in this debate is whether we consider exchange network as one large source of public information with multiple access points and accept non-immediate distribution of information through the network as unavoidable technical detail.
At that point the HFT is guessing and hoping the second order will actually arrive at exchange B. That's the risk the HFT takes.
There is no automatic overflow function that will transfer (the rest of) an order between two exchanges. At best either the original investor can decide to split up his order, or someone else can decide to buy at B and sell at A. But still the HFT will be hoping that actually happens.
The time between exchange A receiving original order (point 2.) and exchange B receiving remains of that order (point 4.) is extremely small, but larger than zero. The argument here is that until exchange B acknowledges reception of the original order in its ledger, that information is not public at exchange B and acting upon that order in exchange B until exchange B acknowledges said order is acting on non-public information, therefore front-running.
> <...> the order flow is technically “public” once broadcasted by the exchange.
The core argument I see in this debate is whether we consider exchange network as one large source of public information with multiple access points and accept non-immediate distribution of information through the network as unavoidable technical detail.