First of, the economy isn't a machine, it's organic, and this engineering approach generally fails. Companies react to regulation, and regulation itself is the result of government, another organic entity. When this type of laws are enacted, the first thing that happens is that concentrated business interest will make sure they actually benefit from the regulation. It can take many forms. Maybe some corporations will be grandfathered in and therefore manage to keep at bay competitors who can't reach a competitive size, maybe the law will have exemptions that only politically connected firms can obtain. It's misguided to push for a law without taking into account the way it will be distorted by the political process. Contrast this to the viral - hence organic - approach the GPL took.
Second, too big to fail is about the systemic risk that some financial firms exhibited. Walmart is big, Google is big, but they're not too big to fail in the sense that their failure wouldn't particularly cause havoc. If Walmart fails, many different sellers can buy the stores and keep supplying them with goods. In the case of financial companies, the argument went as follow: if a bank fails, many other financial companies may be in trouble if they hold financial instruments whose collateral ultimately is guaranteed by that bank. Unfortunately, it can take a long time to sort out who is really it, and during that time, it becomes very risky to lend to anyone, for fear that they might be exposed to the failing institution. This in turns cause more financial companies to fail in a domino effect. That's the theory at least. I don't know if I buy it, but at any rate, it makes the case that the banks were too heavily interconnected to fail, not too big. Columbia professor Rama Cont has suggested that the solution to this problem is to emphasize clearing houses to bring in transparency in who holds what.