Too big to fail? Or too wired too fail?
Simon Johnson (and some sane central banksters*) diagnose the first; Yves diagnoses the second.
For my money -- BWAH-HA-HA-HA-HA-HA-HA-HA-HA! -- Yves has the right of it:
The problem is that we have had a thirty year growth in securitization. A lot of activities that were once done strictly on bank balance sheets are merely originated by banks and are sold into capital markets. Think of the old model as mainframes and the new model as distributed computing.
We have now seen a lot of activity shift from banks to capital markets. And thanks to a host of factors (barriers to entry like high minimum scale, network effects, deregulation which made it easier for firms to span product and geographic markets) we now have capital markets dominated by a very small number of players. And these players are too big to fail by virtue of their ROLE, not simply their size. Lehman was considered small enough to be let go, but it was sufficiently tied to the grid so as to produce a more disruptive outcome than the authorities assumed ...
So the industry had already become so concentrated (and levered) that it had become more failure prone. So merely separating commercial banking and investment banking is not sufficient; you have to do something about the risk taking of capital market players. And sadly, the network effects in trading are powerful. Left to their own devices, the propensity is for the industry to coalesce into a format of fewer, more powerful players. And now that it is in that format, it would take a lot of intervention (Tobin taxes? barriers between products? barriers between geographies?) to not merely restructure the industry, but to keep the factors that favor concentration from reasserting themselves.
Power laws wherever you look.
NOTE * And by "sane" I mean "not completely crazed with greed for money and lust for power."