Update

[Roger Dingledine <arma@mit.edu>]

Ok, I haven't heard from the 'big players' lately, so I feel kind of
out of touch. Can I get a show of hands from everybody on the list, as
to who's in which category?

1. "I think nntp/news is the wrong solution. You're crazy."

2. "I think nntp sounds like a good plan; I'm hoping we can find a
   good solution that we all like."

3. "I'd like to help in this design phase, but I don't know anything
   about nntp."

4. "I'm too busy to help out at all, but I'm interested in participating
   in the 'linux news network' if you guys come up with something useful."

(Or feel free to make up your own categories if these don't fit you. I'm
in category 2, currently.)

--------And now, the actual update

Ok, so it looks like if we're going to do things according to the
current plan, we have two options:
1) Convince the INN developers that they should produce a 'light'
   version of INN: pull out 75% of the functionality, pull out the
   ram caching, etc. This should be done anyway, because then many
   more people would use INN. However, this is not feasible in the
   short to mid term.
2) Bite the bullet and install INN on the news sites, with a custom
   config that we come up with. People have already expressed several
   reservations about this.

There are some other options, though:
3) Develop our own program, based on nntp or some other protocol,
   for doing it ourselves. The upsides of this are that we know it
   all ourselves, and we design it for ourselves. The downsides are
   that we have to actually design and write it, and that there aren't
   other users out there to test/develop it.

4) Use usenet itself. Grab comp.os.linux.news (or whatever), and let
   machines that are already being used as news servers serve us too.
   It appears that these days, messages travel around the world to the
   main news servers in "a few minutes", which makes this feasible (we
   should get some better figures before actually committing to this, of
   course :) It could be moderated, we could post in our XML format (or
   whatever), and the news sites might use suck or some other proxy to
   buffer/spool reads and posts. In short, take advantage of existing
   infrastructure rather than existing software. (There are many more
   details to be worked out for this solution, such as whether we want to
   actually create a hierarchy or just use a single newsgroup, and why
   cola hasn't sufficiently implemented this (some answers were in my
   previous post re: cola, ask me further if you think I'm wrong and cola
   is sufficient.))

5) Forget this concept of networking the news sites. We've gotten quite
   far on creating a "universal language" for describing news articles.
   Let's focus on finishing that and actually implementing it, then we can
   hack together something easy so we can start actually sharing.

Thoughts? Opinions?

Thanks,
--Roger