brad@looking.on.ca (Brad Templeton) (05/18/91)
With some sadness I must report that we have not had success with the trial hierarchy. The busiest group, trial.soc.culture.italian, recently came up for measurement, and with only 62 readers on 23% of the arbitron sites, it ranks only in the 5th percentile of USENET groups by readers/site. We had set a criterion of 25th percentile or better. It fails this test. The members are starting an old-style newsgroup vote, which they will probably pass, since it isn't hard to get 100 votes for anything with the current size of USENET. (The vote system no longer measures support, but rather opposition, in most cases.) We'll see that result at some point. There are several possible factors to consider: a) Very few people have tried trial groups. I attribute a lot of this to Gene Spafford's inclusion of a "you might not get your group anyway, even if it passes" disclaimer on the front of the trial rules. I begged him to remove it, as I think it is unfair. That disclaimer applies to all the rule systems, and in fact, applies most strongly to the 100 vote system, where if there is enough controversy about your vote, the group does not get created, or gets fragmented. Suffice to say that with such a disclaimer in front of a system, I could see why people would not use it. Thanks, Gene. b) Arbitron stats are becoming less reliable. According to them, between 25 and 50% of arbitron sites got trial groups. 25% of modern USENET is as big as all of USENET a couple of years ago. *Plenty* of sites to support any newsgroup. But a couple of the groups fizzled for lack of participation. USENET has increased an order of magnitude while the number of arbitron sites stayed constant, or even dropped. Not good. I had hoped that the trial hierarchy would encourage reporting, as people joined up to push "their" groups. I suspect trial propagation is well under 25% but how can I know? c) I could be totally wrong and people actually like the duke-it-out voting system. In my opinion, that system is getting worse. It's a lot of bureaucracy and noise, and I am not sure how well it is doing in judging groups. d) The groups we got were, no offense intended, not big dynamic winners. A couple of culture groups, a metalworking group, and a software legal issues group that has overlap with many other groups. Perhaps they were intended to fail -- perhaps they never would reach that 25th percentile if everything else were right. ------------------ Courses of action: a) Abandon the trial hierarchy b) Fix it somehow: 1) Lower the criterion perhaps to 15th percentile. (I still think 5th is too low) 2) Change the name to a set of trial hierarchies, ie. comp.trial, soc.trial, rec.trial etc. This would eliminate the propagation question since it would passive to subscribe and active to not subscribe rather than the other way around. (Not subscribing would still be the fairly simple inclusion of a !all.trial line in the sys file) Doing the above (2) would probably require holding a traditional vote, even though hierarchies are not really covered by such votes. If it would get Gene to remove that disclaimer he tacked on the rules, it would be worth it. Comments are welcome. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
jbuck@janus.Berkeley.EDU (Joe Buck) (05/18/91)
In article <1991May17.170212.5145@looking.on.ca> brad@looking.on.ca (Brad Templeton) writes: >With some sadness I must report that we have not had success with the >trial hierarchy. Brad, "trial" was a worthy experiment but almost no one carried the hierarchy. >There are several possible factors to consider: > >a) Very few people have tried trial groups. I attribute a lot of this to >Gene Spafford's inclusion of a "you might not get your group anyway, even if >it passes" disclaimer on the front of the trial rules. They don't try it because their site doesn't even get the "trial" distribution. >b) Arbitron stats are becoming less reliable. According to them, between >25 and 50% of arbitron sites got trial groups. 25% of modern USENET is as >big as all of USENET a couple of years ago. *Plenty* of sites to support >any newsgroup. But a couple of the groups fizzled for lack of participation. Arbitron stats grossly exaggerate the propagation numbers for nonstandard hierarchies, because sites that get everything are far, far more likely to get arbitron than sites that pick and choose groups. Does anyone REALLY believe that alt.sex.pictures has anywhere NEAR the distribution reported in arbitron? It's a self-selected sample. I doubt if even 10% of sites get "trial". Don't believe me? Try a "sendsys" in a restricted area of the net and check. >USENET has increased an order of magnitude while the number of arbitron sites >stayed constant, or even dropped. Not good. Yes, I think that more and more, the only people running arbitron (with few exceptions) are those for whom running mail and netnews is a fulltime job, or folks with lots of outgoing feeds. Such sites are much more likely to carry groups they have no interest in, just to be able to provide complete feeds. >c) I could be totally wrong and people actually like the duke-it-out voting >system. In my opinion, that system is getting worse. It's a lot of >bureaucracy and noise, and I am not sure how well it is doing in judging >groups. For most groups, there is no "duke-it-out" involved at all. The majority of groups are proposed, discussed a bit, then voted on, with the votes going 10 to 1 in favor of the group in question. Since they are so noncontroversial, the existing procedure works just fine for them, and starting in "trial", where most of the net wouldn't get the group, would be a very bad strategic move. So yes, it appears that the net favors the current system. Sorry, Brad. The bureaucracy and noise only happens for the minority of groups that are noncontroversial. >Courses of action: >a) Abandon the trial hierarchy Do it. It was a worthy experiment, but it failed. But there is another possibility: treat "alt" like "trial". This requires no rule changes; at any time someone can suggest a move, and there can be a discussion and vote. -- -- Joe Buck jbuck@janus.berkeley.edu {uunet,ucbvax}!janus.berkeley.edu!jbuck
smb@ulysses.att.com (Steven Bellovin) (05/18/91)
I can't speak for anyone else, but around here almost no ``trial'' articles have ever arrived. This is from my ``active'' file: trial.test 00011 00011 y trial.rec.metalworking 00052 00052 y trial.newgroups 00006 00003 y Not conducive to readership....
ralphs@seattleu.edu (Ralph Sims) (05/18/91)
smb@ulysses.att.com (Steven Bellovin) writes: > I can't speak for anyone else, but around here almost no ``trial'' > articles have ever arrived. This is from my ``active'' file: So far the trial groups haven't made it into NorthWestNet via FARnet. Does anyone know if a FARnet site is handling them? Our news comes from seattleu.edu through milton.washington.edu by way of dali (Montana). -- The 23:00 News and Mail Service - +1 206 292 9048 - Seattle, WA USA PEP, V.32, V.42 +++ A Waffle Iron, Model 1.64 +++
emv@ox.com (Ed Vielmetti) (05/18/91)
> Gene Spafford's inclusion of a "you might not get your group > anyway" I thought Spaf stopped reading netnews months ago, what's he doing making the rules? Time for some alternative monthly publications and lists that have the modern net in mind. --Ed