That's certainly true and why we watch the trends of usage very closely and we project those terms into the future knowing that's imperfect. What we won't do is build networks based purely on guesses. We certainly see demand for upstream capacity increasing for residential customers, but that increase is slower than the increase in downstream demand growth. In all cases but pure greenfield situations the cost of deploying DSL or DOCSIS is significant less than deploying fiber. Even in greenfield situations PON, which is a asynchronous itself, is much less expensive than active Ethernet. In short synchronous connections cost more to deploy. Doing so without a knowing if or when consumers will actually pay for synchronous connections isn't something we're going to do.