Xref: utzoo news.admin:7884 news.groups:15345 Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!uunet!cs.utexas.edu!sun-barr!rutgers!att!cbnews!wbt From: wbt@cbnews.ATT.COM (William B. Thacker) Newsgroups: news.admin,news.groups Subject: Re: Fixing the unbroken Message-ID: <11984@cbnews.ATT.COM> Date: 5 Dec 89 15:18:09 GMT References: <56804@looking.on.ca> Reply-To: wbt@cbnews.ATT.COM (William B. Thacker,00440,cb,1D211,6148604019) Organization: AT&T Bell Laboratories Lines: 104 In article <56804@looking.on.ca> brad@looking.on.ca (Brad Templeton) writes: >In article <11941@cbnews.ATT.COM> wbt@cbnews.ATT.COM (William B. Thacker,00440,cb,1D211,6148604019) writes: >>In article <55499@looking.on.ca> brad@looking.on.ca (Brad Templeton) writes: >>> >>>But I would like to see the "it isn't broken" advocates come out with >>>some evidence to support the claim that this system *is* working. >> >>Usenet continues to grow. People aren't giving it up in droves. >>*Somebody* sure seems to think it's doing well. > >Do you suggest USENET grows because we have such a nice voting system >to pick groups with? No, I merely indicate that whatever's wrong with the voting system can't be *that* bad, because it still results in a popular product. Of course, this doesn't mean it can't be better; simply that it's not all that bad. And I'm not inclined to change it until something comes along that will clearly solve the few problems the current system has. None of the proposals I've seen yet would do so. > Sorry, but I fail to see how you drew this conclusion. >Anecdotal evidence does say people have left USENET in droves. Apparently, then, for every one who leaves, two more step in to take her place. >Of the >active netters of the early days, only a very few still take an active >role. At places like Usenix and Unix expo I regularly get the chance to >ask people if they read USENET, and I very frequently get the answers, >"used to," "too noisy" or "don't have the time." So the fact that I no longer play my euphonium in a band indicates a fundamental flaw in instrumental music ? (Or perhaps just *my* music 8-) Interests come and go. >>Do you still stick with your earlier claim that a group should average at >>least one reader per site according to arbitron ? I have never agreed >>with this belief. > >Actually, I would probably require 1/2 or 1/3 reader per site, but I >don't know how you can disagree with this, given the assumption of >accurate readership measurement. The alternative is to suggest we >send groups to thousands of machines where nobody reads them. Why not ? I've been through this before, but I'll try again. Let's say I'm the sysadmin on fubar.hal.com, and I, or one of my readers, want to read rec.food.artichokes, a small group with only a few thousand estimated readers. Meanwhile, you're the sysadmin of tarfu.buu.com, and have an interest in talk.glue, another small group. Neither of us in intersted in the other's group. I want r.f.a to get wide propogation, so my reader will get the most use of it. That means, among other things, that I want *you* to propogate it to all the sites that feed from you, as they may also have interested readers. Meanwhile, you would like the same treatment for t.g. I therefore agree to carry and forward t.g if you'll do the same for r.f.a. In return for carrying a group you don't read, you get an improvement in the group you *do* read. Likewise for me. Cooperation for mutual benefit. Where this begins to break down is when, after a few months, you realize that the volume in r.f.a is 20 times that in t.g. You're incurring much higher expenses on my account than I am on yours. So it seems to me that volume per reader is a much better indicator of a group's "worthiness" than is number of readers. >The question >that the guidelines try to answer, don't forget, is, "what groups by >default should go to all machines." It seems to me that, rather, the guidelines are to demonstrate that there is sufficient interest in the group to verify that sites you feed are likely to have interested readers, and enough of them to warrant your carrying of the group. They don't equate. A "backbone" site might feed several thousand sites (a few dozen directly, each of which feed a few dozen, etc). Obviously, these sites are much more likley to find interested readers among their "dependents" than would a site that only feeds two or three machines. And in the extreme, a leaf site obviously has no need to carry anything not read by its readers. Therefore, there's no system of voting that can really say "what groups should by default go to all machines;" that choice is left to the sysadmin. >>>My examination of readership from sites that report it says that this system >>>has done rather poorly. It has created very few groups that have reached the >>>highest levels of readership >> >>A useless statement. It is a mathematical axiom that no more than 5% of >>the sample can be in the top 5%. > >My analysis said it was far worse than that, and in particular, if you >recall, that "alt" groups and groups created against the guidelines >(a "control base" if you will) had as good a track record of getting >good readership. Which tells me that the current guidelines are too restrictive. It does not seem to indicate that we need yet further restrictions. - - - - - - - - valuable coupon - - - - - - - clip and save - - - - - - - - Bill Thacker AT&T Network Systems - Columbus wbt@cbnews.att.com Free the Lagrange 5 ! Brought to you by Super Global Mega Corp .com