This is one I'm glad someone asked about because I thought it was clear and it wasn't: I am not advocating a native-client-only approach. I am aiming for something that uses NNTP as mechanism on the back end, so invested users have better options than "bug the 1.5 guys in a position to fix things to fix things" and interoperability between LW and the diaspora is easy rather than hard.

Inbound users should not have to know or care that NNTP is involved (1.7), for exactly the reason you mention: to the average user, the web is the internet (1.1, first because I expect most people here don't realize it either), and explaining to them why they are wrong is not helpful. I want the answer to the eventual, inevitable question "how do I do X" (where X is not some fundamental operation like "read" or "post") to be "install this app today and be happy," as opposed to "bug person Y for six months and hope they take the time to implement X."

The 'cooperate' action I need has to come from diaspora authors, not inbound new users; they are the ones that would need to be convinced to join the network and use blog software that supports it. Making that cooperate action as close to costless as possible (I would settle for "no harder than setting up your own blog") is a hard problem -- but, I believe, solvable.