One idea is a group blog owned by a non-profit, with a charter, bylaws and governing structure explicitly protect openness.
Instead of a single front page, you have cookie-enabled selection of your default front page, and you have a dynamic menu system that allows you to create and manage your own list of different configurations. Configurations can be done by tagging. My preferred approach uses two types of tagging, one that's completely democratic, created on the fly, one that's regulated by discussion to create heterarchical structures (multiple overlapping hierarchies, such as National-State-Local, Law-Constitution-Bill of Rights-First Amendment, etc.) But this is the sort of thing that should be worked out by the community, not put into the bylaws.
The heterarchical tagging should represent real user interest as well. At some points in the heterarchy, there will be sufficient participation to form an editing collective. People here will function like front page posters. (My Left Wing has a very good model for this. Those with Front Page privilege can front page their own diaries or anyone else's, and there are some guidelines. A more rigid structure may be needed, but I prefer leaving that as open as possible.) Any node below one with an editing collective will inherent its front page. Where there's an editing collective at a lower node, they can winnow stories out by "unpromoting" them. Where there's not, we need some sort of winnowing procedure, since multiple inheretence means that front pages would change faster and faster the farther down you go.
Of course, there are two questions here that need serious consideration: How do you form editing collectives? And how do you more stories up the heterarchy? I'd say that ratings should move stories up the heterarchy, subject to editing collective choices.
And editing collectives? First of all, they should be fluid. Ideally, people would serve for a while, then do something else. Some might prefer doing it once a week every month. Some might want to do a 3-month stint, then take time off. We should think of how to accomodate a wide range of different sorts of participation. The less we make it a scarce resource, the better.
We'd have something like a trusted user status, but it would be crossed with a participation record for heterarchical tags, so that we'd have a pool of people who are familiar enough with a topic area that they are legitimately regarded as stakeholders in that area. These are the people who would constitute the pool from whom the editorial collective would arise. I'd favor something simple, such as self-nomination, and a certain number of supporting votes to add someone to the collective. Kicking people off is a harder call. We need to allow it, to prevent infiltration or other destructive developments. But we need to make it a deliberative process, so that it doesn't get abused to silence challenging points of view.
I would strongly recommend the creation of major channel divisions--which should be subject to regular review. This would facilitate breaking news being separated from more strategic and philosophical discussions, which natural flourish on a different rhythm.
And, of course, we need a well thought out way to dynamically handle blogrolls and other links, to make them appropriate to whatever page view the user selects.
This is an idea of a how a super group blog might be organized--one that could conceivably grow in size and traffic to be considerably larger than DKos is now. But note--one of the important aspects of such a blog would be the fluid subcommunities it would help create, and the links to other blogs--and other websites/resources that would be particular to each of those subcommunities. Thus, this would not just be about building a site, it would be about building a node in a larger infrastructure, and about building webs of community, not just one glob-like community.
Anyway, that's just a rough outline of one idea. I'm sure there are others.
With all due respect to Paul, whose comment is very intelligent, and to Pyrrho, who has expressed similar thoughts, it simply won't work.
The problem is who gets to determine citizenship in the community. The problem is who gets to decide what's appropriate. The problem is control.