4 ms·

Hoping to read in detail later, but very interesting at first glance thanks! Have thought for awhile that many systems would benefit from one or more "guard" c

Hoping to read in detail later, but very interesting at first glance thanks! Have thought for awhile that many systems would benefit from one or more "guard" components for which we have formal proofs that no bad things can happen - e.g. you write an equity trading application in Python, but route all the trades through a checker that ensures you aren't trading too much or too often [1]. [1] See Knight Capital 2012, http://en.wikipedia.org/wiki/Knight_Capital_Group#2012_stock_trading_disruption http://en.wikipedia.org/wiki/Knight_Capital_Group#2012_stock...

taeric 12y ago
First, "no bad things can happen" is a very nebulous requirement. And you eventually just get into a "who watches the watchmen" kind of game. I think the problem with a lot of ideas around making a "safe" algorithm for money matters, is that you basically just wind up defining the game that your "opponents" will now be playing against. That is, if you have your rules in the open, but they don't, then they are the ones controlling the game. Consider, it is relatively easy to make money off of a casino. They just don't let you use the strategies that work. Which is fine if you can just boot off actors that you think aren't "playing well." Not so easy otherwise.