From: Brad Templeton (brad@main.templetons.com)
Date: Mon Sep 20 1999 - 15:51:00 CDT
On Mon, Sep 20, 1999 at 02:40:05PM +0000, Dirk Nimmich wrote:
> Maurizio Codogno wrote:
> > Nowadays, Distribution: is completely useless. Were it for me, I'd delete
> > it, since it won't be used anyway.
> > I find "Distribution: local" extremly useful for testing purposes.
A new semantic was proposed for Distribution which is similar to the
original but fixes the problems.
The concept was to supplant the use of distribution in the control of
outgoing feeds with using it to control incoming feeds. (Outgoing feed
control can still remain but it will always be unreliable over multi hops.)
A news database or transport instead maintains a list of the distributions
which it considers itself to be a "member" of. If articles come that have
a distribution line with no distributions they belong to, the system:
a) Definitely ignores any control messages
b) Either does not store the article locally, or is configured to
include distribution in the overview and have newsreaders do
the membership decision for their readers. Corporates sites
would tend to not store locally, ISPs might tend to store and
expect users to pick the distributions they belong to for
reading.
c) At its option, either forwards the article onward or discards it.
Hubs would tend to forward, leaves and branches would tend to
discard.
Distributions then become manageable (self-joined) subsets of the net.
They can also be used on the outgoing feed but now this is only an
efficiency issue, not a correct behaviour issue.
Thus if I want to issue a Nocem stream, I would create a distribution
and use it on all my messages. Sites not in my distribution would ignore
or discard the articles. Hubs would send them around. Sites that want them
would get them and act on them. Sites that wanted to never even see them
would ask feeds not to feed the distribution in question.
As far as I can see it solves all the problems and makes things work again,
and gives the option of ISPs storing all distributions while readers pick
what they like.