If I had more programming talent to invest developing a fed wiki clone (or if I had money to buy programming talent), the direction I'd want to pursue is a sort of Hyper Card-like journal that benefits from federation, networking, and a wiki approach to life.
I am not saying this is where the core product should go. This would be a more narrowly defined fork.
The reductionist pitch would be "It's like Evernote + Hyper Card + wiki. You capture your experience, but you also connect it and share it. You move from capture, to connection, to community."
This is how it'd work.
You'd keep a running journal/log wiki, and this journal would be the core initial view. The journal/log would also collect unexpected fork-to-local instances (yes, I know there are browser permissions issues here, but we're dreaming). If you couldn't save where you were editing, it'd save to your log wiki.
Instead of calling pages pages, I'd call them cards, to reinforce (as with Hyper Card) that these are short, modular treatments made for quick reorganization. It also helps people understand the multi-page layout, and it gets people past the "but I'm not a writer" problem of "pages".
In the journal, you'd note everything you learned in a day. Stuff you heard on the way to work, such as Loved and Lovely. Things you learned in class. Instructions on how to Find Process Using Your Webcam. To see what that might look like you can go here:
<i>Click to see what an example of the journal of Fedwiki as Memex-Journal might consist of.</i>
From the journal, you'd also be able to push these cards to other sites of yours, as well as use more traditional fedwiki "pull" methods. You'd also be able to jump into those journals and build them out individually.
Farms would always be associated with one person, and one login, and users would log in once.
The drag and drop abilities of fedwiki would be improved to support a broader array of "clippings". One of the browser buttons associated with fedwiki would automatically forward the URL you were on to fedwiki, and create the best guess at a Title and initial paragraph it could: YouTube URL to Video Embed, NY Times article to Title + text selected at time of button press + automatic URL link and attribution. (Ward has already done similar).
For scientists and researchers, the cards would support multiple plugins that allowed them to work with data and visualize it, along the lines of the CSV support now. Scientists would also start by journalling their work and then share it out.
It would be important that fedwiki could either replace Evernote, Pinboard, etc. or be used as a seamless adjunct to them, since people think their real problem is capture (they don't realize the real issue is connection). Therefore fedwiki would have to compete as a capture/annotation product.
In general, people would probably keep their running journals private, as this would make them more free to add anything they thought of during the day. However, they would have the option to push-publish anything in their journal to any public journal on their farm.
Unlike wiki, there would be a lot of orphan pages at any given time, and that would be OK. We don't always know where a thought or discovery will fit. But analysis functions would help people identify these pages from time to time and prompt them to connect them.
Theory and Pitch
I think the pitch for this is simple.
People are capturing all this stuff now -- Evernote, Pinboard, Readability.
People are sharing this stuff across networks -- Pinboard, Pinterest, Facebook, Twitter.
People are even annotating: Diigo, etc.
But we're missing three things.
First -- we aren't connecting these things. We're pushing around atomic units. Even personally, we need to connect, supplement, reflect. Without that, Evernote just becomes a new shoebox of papers.
Second -- We share and capture these things -- blog posts, data, pictures. But we don't make them better. The lesson plan my wife writes for art that she shares on Pinterest doesn't get better as it travels the network. The data I share doesn't get connected to cool visualizations. We need to let the network act on these things to really harness the power of it -- otherwise it's just a routing mechanism. Even personally, everyone has their head too much in the stream, and doesn't revisit old work.
Third -- More debatable, but I'd also say serendipity is lacking in some current systems. This is harder to explain, but worth keeping in mind. We're getting less and less surprised by our networks, and that's bad.
Applications and Markets
Education -- Student journals become networked and meaningful. Entries become building blocks for future work and understanding. The ePortfolio discussion becomes simple -- students share work to a public wiki (card stack?), the institution forks what it needs to its own servers to collect evidence of learning.
Business -- Let business people know that all the stuff they are shoving into Evernote is meaningless without connection. Sell them on the idea that they could quickly rearrange notes into a business plan, and use a single system to share out everything from contact cards to case studies to meeting notes.
"Life-hackers" -- these people use a lot of crappy software to build memory, mind map, increase productivity. Show them that what they really need is a sort of "mind-mapping" evernote and they'll come.
Along with Ward I imagine a Pinboard-ish community around the product. Sites would have a setting to say where they publish to -- RSS feeds, Pinboard, etc. But there also might be a fedwiki specific community that provided better integration.
Wikis would also have certain tags associated with them, and by default would publish new material to feeds and community sites under those tags. Tags would help alert you to new wiki content from anywhere, consistently good wiki content would prompt you to subscribe to all updates of that wiki.
I like the WordPress business model. Open source software plus a service business. That combination has created a situation with WP where 20% of all domains are running WordPress.
That said, I don't have the personality to run a services company. So someone else would have to do this.
The Evernote Model might be interesting. Free software, you get charged based on how much you upload each month to the service. Incentives here point the right direction for both customer and business.
A possible product model is free browser tool, paid mobile app. As always, all is open source.
I think it'd be a derivation of fedwiki, not the core product.
I like the name "Steno". It conveys the notebook idea, but technically stenography is "narrow writing" (steno=narrow) which fits the idea of a collection of small thoughts connected. It doesn't capture the networked wiki element, but I think that's OK -- it's easy to say "Steno is your networked notebook".
I'm not a person of means, and I'm more a PHP/Python hacker than a real coder, so this is idle thinking. But I have this ready to share if I meet someone looking for an opportunity.