Gerard, with all due respect, your reply is all based on incorrect assumptions. I recognize the severe problems that mediawiki conversations currently have, and my points about Flow acknowledge that it's incomplete software at its early stages and that it can grow into an acceptable tool for having discussions, but all that is irrelevant to the conversation that's going on at this thread.
Both Erik and I are talking about what we expect a finished, full featured software would look like in the end, and we have both dismissed the current status of either tool. The idea that I want the new system to be based on current existing software is a strawman; that's not what I defend. I want a good, modern document-centric software even if it requires building something like mediawiki from scratch. This is a high level view of what either model can grow into if enough resources are poured into it.
Erik has this vision that building a stable and easy-to-use system requires abandoning the open-ended nature of wiki systems, with which I disagree, and has committed us to a project with that result in the end, to build a very good architecture that will solve the wrong problem. From the conversation so far, I believe such view to be based on an incomplete understanding of the community needs, and I'm trying to steer the conversation to take into account perspectives from the wider community rather than the gut feelings of just one man, no matter how much experience he has with the project.