Usenet Los Angeles

Background

Ever wonder how Usenet works? Articles go in here, and pop out everywhere, and nobody has to lift a finger to make it happen! Pretty cool, eh?

Except that it does take work. A full Usenet feed is 450 megabytes per day, and the news servers have to be up nearly continuously. If they run out of CPU or disk space, whoops! articles start getting dropped. This happens often; probably several news hosts are down at any one time.

Fortunately, Usenet can work around problem hosts, if alternate feeds are available. For a partial map of how Los Angeles Usenet sites fed each other in 1994, click here. Note how many sites fed several other sites. The whole thing looks like a maze of twisty little connections. Generally, the fewer hops you are from a major (and therefore well-run) site like JPL, UCLA, Cerfnet, or UUNET, the less likely you were to lose messages because someone else's system was misbehaving.

The Problem

Unfortunately, many commercial Usenet sites in Los Angeles have to pay for alternate news feeds, and can't always afford them. Furthermore, many of the feeds come from faraway sites that don't particularly care about la.*. So it seems a lot of messages are getting dropped, delayed, or even duplicated on their way cross town.

On Nov. 30th, 1995, I logged in to three different Usenet sites in LA and looked at la.personals and comp.lang.java. The differences were striking; many recent messages were missing from both groups on two of the sites.

Here are a few notes about a test I ran on news propagation.

What to do?

First, I'm trying to see if I can understand what's going on. Second, there are a few commonsense things we can do:

Usenet groups can't propagate well if people don't agree on what they are. (There has been confusion over la.news, and not every site carries newer groups like la.config.) After some discussion on la.config, I put together a note describing the la.* newsgroups; it's a very mild document. No statements about allowed use are needed to deal with propagation issues, so I'd rather not try to define allowed or disallowed behavior.
Usenet groups can't propagate well through sites that don't know about them, so we should set up an periodically posted checkgroups message to inform new Usenet sites of the la.* group names and descriptions. I can easily do this. (Might use 'addgroups' messages to avoid deleting other people's la.* groups not yet in the list.)
The simplest way to detect la.* newsfeed problems is to watch the weekly Usenet status messages on la.news. If you stop seeing them, there's a problem. If they arrive late, there's a problem.
You can help out other news admins by posting your own weekly messages to la.news. If the messages contain a summary of how many messages were seen in the la.* groups at your site, so much the better, but even an empty message would be quite helpful. Here's the script I use to post stats. (It was written rather quickly, so caveat emptor.)
If you are a Los Angeles usenet site, but your newsfeeds are all out-of-town providers like uunet or sprintnet, arrange to exchange la.* newsgroups with one or two local usenet sites like yourself. Both ends of these tiny newsfeeds benefit equally, so it should be possible to set this up with the other news admins informally. This is the original basis of Usenet, and it's a good idea that still works! (Jet Cafe (jetcafe.org) seems to have people who live and breathe this stuff; they might be able to tell you more.)

I invite comments. Please post any discussion on this topic to la.config - and pray it gets out past those flaky sites.