brad@looking.UUCP (Brad Templeton) (04/10/89)
While there is discussion of the mechanism for figuring out which groups are good, Gene is right that some discussion of the philosophy behind how new groups are evaluated is worthwhile. I have seen two main camps of newsgroup creation philosophy. They both have very strong merits. I have wavered between them myself. The first camp says "Create only truly necessary groups -- keep the 'namespace' small." The second camp says, "Create any reasonable group -- what trouble are small volume groups?" Oddly enough, the Camp 1 philosphy sprang out of old software limitations. There were days on the net when a large number of sites (including old Pdp-11 sites) simply couldn't handle more than 256, and then 512 newsgroups before they died. So 'namespace' had to be preserved. That is gone now, or mostly gone. Today, however, the argument is that an explosion of thousands of groups would simply make the net too confusing to all but the most experienced netters. To me the most compelling argument for Camp 1 is that new newsgroups definitely do cause traffic that would not have otherwise existed. While some groups might mostly split up traffic, it seems clear to me after 8 years on usenet that new groups do cause more traffic than would have existed before. Camp 2 has good arguments as well. Usenet needs more classification of articles to allow people to control their reading. Some groups are simply too large and noisy to read. And to me, high volume is the measure of a bad group, not a good one. There really is no harm from a small volume subgroup. Subgroups (level 3 or more) don't confuse the namespace. And they don't create traffic as much either. (This assumes the group can sustain one reader per long distance link it crosses. Any group that can't do that should not be on usenet) ------------- So what's the answer? Perhaps some sort of mixture is possible. On systems like Compuserve, Genie, Bix, Delphi etc. they divide their groups up into categories which the sysop controls. So while they may only have around 100-200 major goups on such systems, each group has 5 to 15 categories. The total number of discussion areas can be in the thousands without confusion. For example, an "sf-lovers" group would have Star Trek, Doctor Who, Comics, SF Books, SF Movies, Fandom etc. all in different categories under the one roof. -------------- So a possible answer? Set up a procedure (of some kind, either UNLAB or another) for the creation of level 1 and level 2 groups. But for each level 2 group get a single group controller who is free to create subgroups of level 3 and beyond at will. This would not be a moderator who controls messages, just a person who knows the group well and can tell what groups are good ideas for new traffic. This does not confuse the namespace, and takes simple decisions out of the hands of a complex procedure. At my suggestion, Eric Raymond has put a feature into his TMN news to support this concept. Under that system if a message comes in to rec.arts.sf-lovers.asmiov, and the host does not have that group, but does have rec.arts.sf-lovers, then it goes into the first higher level group found, or junk beyond a certain level. (This is how you finally did it, right Eric?) In this way, sites that don't want expanded subgroup namespace don't need to have it. Sites that want it get to have it. Of course, we still need a mechanism for level 2 and (rarely) level 1. I still suggest UNLAB for level 2, and possibly participatory democracy of the sysadmins for level 1. -- Brad Templeton, Looking Glass Software Ltd. -- Waterloo, Ontario 519/884-7473
csu@alembic.UUCP (Dave Mack) (04/12/89)
In article <3065@looking.UUCP> brad@looking.UUCP (Brad Templeton) writes: >I have seen two main camps of newsgroup creation philosophy. They >both have very strong merits. I have wavered between them myself. > >The first camp says "Create only truly necessary groups -- keep the >'namespace' small." > >The second camp says, "Create any reasonable group -- what trouble are >small volume groups?" > >To me the most compelling argument for Camp 1 is that new newsgroups >definitely do cause traffic that would not have otherwise existed. While >some groups might mostly split up traffic, it seems clear to me after 8 >years on usenet that new groups do cause more traffic than would have >existed before. Perhaps some, but how much? Consider: there are N people who post to the net and each of those people only has a certain amount of time in which to post. If they distribute that time across more newsgroups, then either they don't post to groups they previously posted to or they post shorter articles. (One could argue that they spend more time posting, but that has an obvious limit; if they spend too much time posting, they flunk out of school or lose their job or starve to death, and thus cease posting altogether.) Obviously, N is growing with time, but that isn't necessarily a result of increasing the number of groups. The increasing volume on the net is a mainly a function of the number of people on the net, not the number of groups. The "Camp 1" argument is fallacious in that ignores the fact that people only have so much time to read and post. >Camp 2 has good arguments as well. Usenet needs more classification >of articles to allow people to control their reading. Some groups >are simply too large and noisy to read. And to me, high volume is the >measure of a bad group, not a good one. [...] >At my suggestion, Eric Raymond has put a feature into his TMN news to support >this concept. Under that system if a message comes in to >rec.arts.sf-lovers.asmiov, and the host does not have that group, but does >have rec.arts.sf-lovers, then it goes into the first higher level group >found, or junk beyond a certain level. (This is how you finally did it, >right Eric?) I think this is a step in the right direction, but it doesn't solve the problem of users finding the groups that interest them. How do I know to ask for "rec.arts.sf-lovers.asmiov" (asmiov - heh!)? Part of the argument between the two "camps" stems from the fact that newsgroup name is used for two functions: it specifies the general area of discussion for messages in the newsgroup and it is also used to specify the way in which the news is distributed. One is a user interface function, the other an administrative function. While some degree of overlap is clearly necessary, it seems that the two functions should be more separated. What we need is a method of extracting articles that interest the individual reader independent of the name space. Example: if I want to read articles about the cold fusion research going on now, I have to subscribe to alt.fusion, sci.physics, and possibly misc.headlines, and I'll still miss articles posted to groups like sci.chem that may be relevant. What I'd like is an interface that allows me to type "/fusion | deuterium/{sci,alt,misc,talk}" and returns a list of articles from those top-level groups that contain either of the words "fusion" or "deuterium" somewhere in them. Obviously, grepping through every article is impractical. Maybe what we need is a newsposter that automatically fills in the Keywords: line at the time the article is posted, combined with a newsreader that searches for specified tokens on the Keywords line of all articles in specified groups. There are problems with this approach, of course, but I suspect that this is the direction we should be heading. -- Dave Mack