Re: Re WIT

Brian Behlendorf (brian@wired.com)
Thu, 9 Jun 1994 14:52:54 +0000 (GMT)


On Thu, 9 Jun 1994, Terry Allen wrote:
> >> I agree, the idea is good, but the implementation is too clumsy
> > to be practical if we can use email instead.
> It is a little bit harder to use but what is going to happen when we have
> 1500 email messages on this topic? Who is going to organize all the
> thoughts so that we could ever hope to go back over it all and apply the
> information over the long term.
>
> In response, I have to remark that only a human could organize the
> *content* of 1500 messages anyway, and WIT doesn't improve on
> properly labelled email in that regard.

A few things that will help this:

1) A "gardener" who prunes the tree as needed - either by directing
questions appropriately, taking sub-sub-sub-topics and returning them to
the main level when they've deviated sufficiently from the main topic,
etc. It'll be a while, at least it seems to me, before the sculpting of
the information can be automated and managed to the point where it can be
done by an intelligent agent or script. Moderation, if done correctly,
can be a Very Good Thing.

2) Seal of Approvals - instead of being able to merely "agree" and
"disagree", users can apply "approvals" and "disapprovals" to
any post they read. No comment to be left, just a general sign of
support or an indication by the reader that this is a Good post.
Then, users can index information according to how many seals have
been applied (per unit of time, perhaps), and arbitrary users's
approvals could be given arbitrary weights. I.e., I could ask the
server to organize the threads according to how well liked that
thread was by Tim Berners-Lee, or how well disliked they were by
David Sternlight.

None of these, of course, solves the central issue, which is one of
scalability. Maybe we're attacking this issue wrong - maybe it'd be
better to hack on the NNTP protocol and clients to allow tree-like threads
and hypertext links everywhere. On the other hand, is the paradigm of
20000 newsgroups distributed uniformly over 20000 machines necessarily
better than 20000 newsgroups, each on one machine to itself? Do
caches make this question moot?

Interesting times ahead.

Brian