Infact, if I was an ISP I would be seriously considering NATing my entire customer base - except for those who were willing to pay $x/mo for a real IP. At the end of the day, with a few exceptions, most protocols work quite well from behind a nat.
I hope your helpdesk is well staffed. The only customers you're likely to get are the ones previously noted as the type who 'wont care'. That being those without much in the way of a desire to use anything that NAT breaks. Your helpdesk will need to cater for the fact that the average level of IT-Cloo of your customers is likely to be lower; and the overhead of having to explain why xyz application doesn't work properly without more money being thrown at their internet account...
And it gives me another revenue stream, which if I am an ISP would be quite useful considering I am probably losing money on the DSL I am currently providing.
It wouldn't suprise me if customers took the opportunity to consider another ISP; I know many that do so when looking to make significant plan changes... To implement this sort of thing seems like a gamble that is, at this stage at the game, no better than revenue-neutral. Maybe when IP address space is a more scare commodity, it'll look like a better option....? Mark.