I started to type something similar then did some research into the costs. The DB download is 8.8GB, they say it expands to terrabytes when uncompressed. That's no longer an impossibly difficult amount of storage to acquire though. The Wikimedia Foundation budget for this next year is $42.1 million, I only skimmed so I didn't find a breakdown on costs of things like electricity and bandwidth out of that. I like the idea, though wonder how you could gain traction from potential contributors. Would those alienated by the current Wikipedia processes along with some other people interested in the novelty be enough to start a community?
The actual infrastructure would probably be relatively cheap (my el-cheapo Hetzner hosting plan has 3TB of space for instance), bandwidth is another matter entirely - such a project would have to be ad supported I think, it's the only thing that scales with the visitor count.
Community is a big issue, but it wouldn't necessarily need to be large, especially at first a small band of moderators would be enough. Initially, such a project would base its data off Wikipedia entirely, but over time more things need to be curated. While we're on it, voting on articles and moving discussion threads about them to the forefront might also be worth looking into. Social proof could be one of the mechanisms used to determine the "best" version of an article. I think there are hundreds of parameters worth experimenting with.
I guess there are a lot of people alienated by Wikipedia, but the beauty of the idea is that we wouldn't have to rely on that. Incoming links would in time drive most of the traffic to the site, as it would be - by definition - a better place to link to for durable and extensive information.
Sounds like a nice idea for a quirky, bootstrapped startup. :)
I like the idea, I wonder if it has to be ad revenue though. I find that idea (personally) distasteful, but don't have any better ideas besides getting donations.
Maybe this exists, MediaWiki may even support it. What I'd like to see is a wiki that allowed you to populate articles by crossreferencing other pages.
An example:
Page: Lisp <tagged as FP>
A functional programming language invented in...
<Portion marked as summary>
Page: Functional Programming
<Summary of FP>
List of languages in the FP category:
<Populated from tagged articles>
<Populated from summary in tagged articles>
You'd only edit that Lisp summary in one location, then other articles could pull from it as needed. Not sure how well it'd work in practice, and as I said maybe MediaWiki and others support this now. I haven't looked at their features in years.
If you initially constructed your wiki from Wikipedia (and maybe other sources), the initial effort would be on removing this redundancy, but you'd still provide as many articles they'd just be better synchronized with each other. Adding breadth (Scheme, Common Lisp, etc) would be straightforward by sharing whole blocks of text. Shorter articles wouldn't be penalized because they'd just be pulled into others via this same cross referencing mechanism. They could be hidden in some fashion to avoid namespace clutter, but outright deletion would be unnecessary.
Well that was a braindump, and maybe not as coherent as I thought it would be. This is why I should sleep more, and practice describing my ideas to other people for feedback.
EDIT: To continue, a thought I should've included originally, use some form of markup for those pages to autofill parts and maybe autogenerate the summary. A language has a place, person and time of invention. A set of paradigms it may be described as fitting into, a set of parent, children and cousin languages. Marked up, this allows for portions to be very consistent in presentation across pages in a category. I suppose I should look at semantic wikis again. As I said, this was an idea I had years ago and pondered but never really researched.
