brad@looking.on.ca (Brad Templeton) (11/13/89)
This debate has grown more intense, so I thought I would summarize my thoughts in one article. A lot of people are missing the forest for the trees here. This is my impression of the forest. What is the purpose of a group creation process? To create a group, one must generate some sort of objective demonstration to net sysadmins that the group will be valuable and in demand. Believe it or not, that is the essence of the 'newsgroup guidelines' in one short sentence. The current 'guidelines' are just one method of providing such a demonstration -- take a survey. But people forget that they are just one method. What was the *real* purpose of the guidelines? I lied. The guidelines were written only partly to satisfy the above criterion. They were actually written to provide a definition of when group creation *failed* more than to define when it was ok. In other words, they are there to get a "group champion" to shut up when he or she loses. They do this through the illusion of democracy, since it is ingrained into our souls never to question democracy. They are, however, now failing at their secret goal. What's in a name? Names are organized into hierarchies to make the system easier to understand, and to make it easy to find groups, and select and subscribe to them by general as well as specific topics. A namespace should be clear, somewhat intuitive and above all, consistent. What's really in a name? The "great renaming" had another goal which has come to surpass the above goal. Groups were organized into hierarchies because of the limited nature of the B news "sys" file for site subscriptions. Renamers attempted to group groups according to value jugdements of their utility to computer corporate and educational sites. Groups were put in "talk" and "rec" under the theory that it would be easy to shut the whole hierarchy off as desired. This contradicted the "great renaming"'s supposed goal of splitting distribution and group name. (In the old days, all groups had their distribution prepended.) This has failed, as now people regularly try to name their groups to get good distribution, not according to their subject matter. They always will do this, as long as the name defines the distribution. What does a 'vote' mean? Usenet group votes attempt to measure interest in a group (YES votes) and at the same time opposition to the group (NO votes). Disinterest in the group is not measured. If a group has a fair bit more interest than opposition, the theory goes, that's an objective demonstration of a valuable and desired group. What do 'votes' really mean? Net votes are really self-selected surveys, something statisticians will tell you are of remarkably little value as an objective indicator of opinion. In addition, the two values (YES & NO) measure two almost orthoganal things. The authors of the guidelines understood the self-selectivity to a degree by asking for a 100 vote margin. But it's also clear that anybody with the energy can alter vote totals tremendously if desired. All objectivity is gone -- that's why the statisticians say what they do about self-selection. Indeed, it is the case today that a 'vote' result depends more on the actions, skill and enthusiasm of the group champion than on the topic in question. In addition, at least on arbitron sites, 'vote'-created groups have a dismal readership record. Very few groups created under these guidelines are widely read on those sites. Should groups be easy or hard to create? This is one of the oldest debates in news.*. There are two large camps for either side. So far the 'hard to create' camp has held the day. I believe that the answer lies somewhere between the camps. Groups should be easy to create within established hierarchies, and new hierarchies should be harder to create. But I won't resolve this issue in this essay. It's worth a whole essay on its own. Why does group creation have so many flame wars around it? Hard to say, exactly. Probably because creating groups is the closest thing to 'administering usenet' that there is, and everybody wants to run the net. But also because we invite them. In a minarchy like USENET, rules are no good if people constantly argue about interpretation. Even worse if people feel it's worth debating tiny technical details. People love to pretend they are in authority, and love to appoint themselves net.police when they see a violation of the current guidelines. This serves no end. The simpler the process, and the less room for debate or 'enforcement,' the better. How can we pick names? Even if you consider the secondary purpose of the great renaming, it is clear to me that it is most important that the naming of groups be consistent. More important than that it appear democratic. Picking names through any democratic process (such as name votes of any kind) is sure to make an inconsistent system. As is having the group champions pick names. Even if the picking of *topics* that are worthy of groups remains democratic or pseudo-democratic, the picking of names should be disassociated from the selection of topics. To my mind, any competent experienced netter could pick names well for groups that he or she doesn't have a personal interest in. It just isn't that hard a task. In fact, if nobody else will do it I would volunteer. But I think it should not be myself, Greg Woods, Chuq V. Rospach or anybody else who has a real or apparent axe to grind on group creation. Any lurkers? Ideally, however, we have to split off distribution from naming before name picking can become the simple, non-controversial task it should be. How can we pick groups? I've said it before, but it sure has taken a lot of evidence to get people to believe it. The guideline system isn't working. It was created to find an objective measure of good groups and to reduce flaming and noise. It is failing at both these goals. My current suggestion on how to pick groups was posted recently to news.groups (Subject: So you want one that fits in 25 lines?) In the long run, however, I support a dynamic mechanism that allows arbitrary creation of discussion topics within groups -- probably somethink akin to the sub-categories found on all the commercial services these days, and notesfiles. Indeed, I could envision a system were all groups start on one system, and grow outwards as people hook up and get them. Where no group is netwide until people netwide start participating. But a lot of new software will have to be written and a lot of packets will have to flow through the bridge before that happens. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
peter@ficc.uu.net (Peter da Silva) (11/14/89)
Brad: there is one big problem with your article. It implies that a technological fix to the social problem would be easy: getting rid of the "B News sys file". I disagree. The problem is not the sys file... it can be made arbitrarily large with very little effort. The problem is administrative. No matter what the format of the file that controls groups, it's just too much trouble or keep track of every new group someone comes up with. I currently maintain a nearly per-group sys file at Ferranti. Each group or level gets its own line, and then it's automatically crunched for News' pleasure ... it's as easy to administer as any per-group system would be. And I'm about 3 weeks behind in updating it. It's just too much hassle to track all the groups. But you need to have some sort of explicit filtering. Your .newsrc-based filter plan sounds good, but it would lead to unintended distribution problems. This is why I suggest things like: moving the personal computer groups out of comp.sys to comp.pc, standardising the hierarchy, making the first level groups well chosen. This is why I railed against comp.soft-sys. And you're trying to put *everyone* through the same wringer? -- `-_-' Peter da Silva <peter@ficc.uu.net> <peter@sugar.hackercorp.com>. 'U` -------------- +1 713 274 5180. "*Real* wizards don't whine about how they paid their dues" -- Quentin Johnson quent@atanasoff.cs.iastate.edu