I think it will likely fail in either way.
Connecting to many networks has a problem that networks are different and if you implement a common protocol for all the networks you are using only subset of their possibilities thus using original client yields always better results. Second problem is that you are not able to keep up with development of all the networks. Protocols are changing, new and better features are rolled out all the time and so on. You'll need development resources of all the networks you are supporting COMBINED plus resources that go to reverse engineering and understanding closed-source protocols
Making your own network will fail because there are already too many networks thus it is very hard to attract critical mass of sharers that is needed for network to take off. These days it is not enough to make a program that is better because most of the current networks are already quite advanced and problem is not network technology or client program but rather availability of resources, most importantly upload bandwidth. And problem is getting worse constantly. Main source of the problem is current broadband scheme where download speed is 4 or more times faster than upload. That means that in the future networks are not competing with technology so much, but rather with available upload bandwidth. In the future technology doesn't matter so much but rather business aspects such as making deals with cable companies to pre-install P2P program to all clients and so on.
It is possible to make better P2P network than current ones, but it won't be 2 times better, but maybe 10% better and this margin is shrinking constantly. In few months it might be 5% already. And in my oppinion it is not enough to attract users to switch over, considering that initially there will be no content
Of course all above is only my oppinion and might be utterly wrong
If you are still going ahead with development, good luck to you
