[ntp:questions] Re: Leap second talks are postponed
janspam.ceuleers at computer.org
Mon Nov 21 19:31:01 UTC 2005
> On Sun, 20 Nov 2005 20:33:45 -0500, "Richard B. Gilbert"
> <rgilbert88 at comcast.net> dropped the following oil-slick:
>>I believe that the requirement, in the US, for time stamping financial
>>transactions is +/- 2 seconds. Even if a system fails to handle a leap
>>second properly, it should still be within specification.
> Robust systems do not rely on the timestamp alone to uniquely identify
> a transaction and can recover from situations where the time can be
> out be several minutes,
The transaction time-stamping requirement is not about robustness (i.e.
certainty that the transaction is executed correctly), but rather about
the regulatory requirement of so-called "best execution". This requires
financial institutions to be able to demonstrate that when executing
market trades on behalf of their clients, they do so in a way that best
suits their clients rather than doing so in their own interests. A
blatant contravention of this principle would be where a financial
institution exploits the information that is implied in a client's
market trade order for its own purposes (e.g. by performing a few trades
for its own account) before executing the client's trade.
The order in which trades are executed is therefore important, as is the
time taken to execute a transaction.
More information about the questions