david@gorm.ruc.dk (David Stodolsky) (05/30/91)
>In article <1990Nov6.203753.26978@daimi.aau.dk> dsstodol@daimi.aau.dk (David S. Stodolsky) writes: >>A simpler explanation is that a trial simply does not measure the same thing as >>a vote. A vote takes an absolute measure of interest in a topic. Trial measures >>a relative interest, either in comparison with other groups or in number of >>readers per machine. Thus the trial proposal is basically flawed (if it attempts >>another approach to current criteria for group creation). And flawed in a way >>which inhibits growth of the net. > >Interesting supposition. I view this as a feature, however, and not a bug. >It would have been easy to define an absolute standard of readership, or >more appropriate, readers per machine. I felt that a relative standard >was more appropriate, as it would change as the net changed. > >I think the absolute standard of yes - no > 100 (with the recent addition >of && yes > 2 * no) is more flawed as it has failed to adjust to >a much larger net. > >-- >Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884- 7473 If we follow this logic, then when the Net is 10 times as large we should require 1000 yes votes to form a newsgroup. Since the current 100 votes requirement is based upon a mailing list vs. newsgroup trade-off, it is easy to see that trial.* is flawed - if it attempts to maintain reasonable work loads for mailing list managers (not to mention drowning the net in bounced mail). Do we want to have a rule that says "it is permitted for 100 people to organize a newsgroup" today, but tomorrow when the Net is bigger and has more resources the same people can not do the same thing? Seems to me, things should go the other way. Usenet should be the place where any group that is excluded by other media should be able to come and find a way to communicate effectively. Trial.* could play a role here, where anybody could kick off a trial balloon (no pun intended) and see how much interest there was. This assumes we want Usenet to grow. An lets face it, if Usenet doesn't play this role, some other net will. What causes the difference in view is the unit of analysis we start with, and relative vs. absolute measurement. You base your analysis on number of machines on the Net. I base my analysis on the number of people interested in a subject. Since the machine to person ratio is constantly changing, the relationship between the views will also change. In the limit each person will have their own machine. In that case, your view leads to a competition between different topics for propagation. My view leads to creation of any newsgroup that has adequate readership. Your view is based upon conservation of machine resources. Mine is based upon the conservation of human resources. You have pointed out a number of times that it is the human resources that are costly. Drop trial.* or rewrite it so that it measures the right thing. Another weakness of the trial proposal is its dependence upon an elite group to select names. The idea here, I guess is that it is faster than a vote. Currently, creating a group takes a few weeks of discussion to select a name or names, a few weeks for the vote, and a few weeks for the newgroup committee to issue the newgroup command. This is a unique feature of Usenet, it take three people the same amount of time to reach a decsion as three hundred. That being the case, there is a better chance of reaching a good decision with the larger group. (Sure, this is not true for the new enthusiastic trial.* judges, but once trial.* becomes established and bureaucratic inertia sets in, the same dynamic is likely to appear.) Your mission, should you choose to accept it: How to get the right name from the net at large. (maybe preferential voting :-) -- David S. Stodolsky Messages: + 45 46 75 77 11 x 24 41 Department of Computer Science Tel: + 45 31 95 92 82 Bldg. 20.1, Roskilde University Center Internet: david@ruc.dk Post Box 260, DK-4000 Roskilde, Denmark Fax: + 45 46 75 74 01
emv@msen.com (Ed Vielmetti) (05/31/91)
trial.* has already been reformed. Its function has been taken over by (surprise!) alt.*, in particular newsgroups like alt.comp.compression alt.sci.astro.fits alt.comp.acad-freedom.talk These groups were created in alt with names which exactly reflect their eventual names in ``mainstream usenet''. Traffic flows, people get an idea of what the eventual group charter will be once there's enough interest gathered together to go through the rigor of a full vote, and incompletely specified ideas get lost in the morass of alt. (no great loss there). If the name loses, the group never gets put to vote. The voting rules are still useful -- given the incomplete distribution of alt.*, the measurement of when it's reasonable to migrate to ``mainstream usenet'' is a show of hands, and using a vote as a show of hands is as good as any. Many of the yes voters will be people who don't have access to alt.* because of resource constraints at their site. This analysis, like all analyses of alt.* behavior, is strictly my own opinions; there is no monthly What is the Altnet monthly posting with hoary interpretations of the lore and collected wisdom of alt. Just as well. It does suggest that people newgrouping alt groups should take care to pick a good name if they eventually want a ``mainstream usenet'' group. So picking a name alt.fax is not as good as what could have been done in retrospect as alt.comp.dcom.fax since the eventual vote would have been trivial to run. There are alt groups which belong in alt forever -- alt.sex, alt.drugs, alt.security come to mind -- which have an alternative manner about them. I don't expect these groups to be renamed with dumb stupid five-part disjointed names to satisfy organization freaks. -- Edward Vielmetti, MSEN Inc. moderator, comp.archives emv@msen.com "often those with the power to appoint will be on one side of a controversial issue and find it convenient to use their opponent's momentary stridency as a pretext to squelch them"
vancleef@iastate.edu (Van Cleef Henry H) (05/31/91)
Our experience with trial.soc.culture.italian indicates that the trial hierarchy has some fundamental problems. 1. The trial hierarchy is not carried at too many sites. 2. The net backbone does not propogate the news entries because of "holes" in the propogation. We are presently conducting a vote for creation of soc.culture.italian. The votes are being taken at another site, and I do not know what the status is, but based on E-mail discussions I would expect the vote to end up about the same as for the other soc.culture groups created recently. Use of the alt hierarchy for "testing out" the concept of a group like this has a fundamental propogation problem in that many sites do not carry this hierarchy. Some of our support for this group comes from commercial sites in Italy---I do not know if these carry alt, but it is reasonable to assume that their managements, like US company managements, would consider alt to be neither work-related nor culture-related, and not carry them. I know of private corporate networks who do carry the equivalent of soc.culture.* groups as a part of their corporate culture. My earlier comment that I saw such groups for various countries as things that "ought to be" was, unfortunately, misunderstood, and I will reiterate, we are going through this vote 100% strictly by the numbers. I do feel that in the long run the net community needs to consider methods other than the voting process that can be used to get the idea of a particular newsgroup in front of the people who would be reached by it. I have not looked at the arbitron statistics for the readership of this group, but suspect that even the voting process, which seems to reach more potential -- Hank van Cleef vancleef@iastate.edu Iowa State University, Ames. Ia. tmn!vancleef The Union Institute, Cincinnati, Oh.
brad@looking.on.ca (Brad Templeton) (05/31/91)
The answer to the question of the right criterion is uncertain. What is truly decided if a group is to be created? Really only one thing; that the group will, by default, get created on all 20,000 or so USENET machines. It only sets the default, as of course anybody can elect not to have a group. And anybody can elect to have a group without a vote if the admins at sites involved are convinced to create it. This means very few sites. It is not efficient in human or machine resources for a discussion group of 100 people to have a discussion distributed to 20,000 machines. In the future, I think that something along the lines of the dynamic feeding work I have been doing will have to become the norm. USENET is now at a size that almost any topic you can imagine has well over 100 people interested in it. Indeed, my perception is that the votes have now become measures of opposition to a group. Fewer and fewer groups fail to get 100 interested parties. Any group with an energetic champion is assured those 100 parties. Instead, groups are defeated by opposition. What an odd system. If the question is indeed, "is there enough interest that the mailing list would be too big" then 100 yes votes (with no votes simply not counted at all) is possibly the right answer. If the question is "should this group be created by default on all machines" then we need to measure broadbased support. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
david@gorm.ruc.dk (David Stodolsky) (06/01/91)
brad@looking.on.ca (Brad Templeton) writes: >The answer to the question of the right criterion is uncertain. How do we choose the right criterion? >It is not efficient in human or machine resources for a discussion group >of 100 people to have a discussion distributed to 20,000 machines. In the Support this statement - if you can. -- David S. Stodolsky Messages: + 45 46 75 77 11 x 24 41 Department of Computer Science Tel: + 45 31 95 92 82 Bldg. 20.1, Roskilde University Center Internet: david@ruc.dk Post Box 260, DK-4000 Roskilde, Denmark Fax: + 45 46 75 74 01
gl8f@astsun.astro.Virginia.EDU (Greg Lindahl) (06/01/91)
In article <1991May31.192813.28009@gorm.ruc.dk> david@gorm.ruc.dk (David Stodolsky) writes: >How do we choose the right criterion? By using our heads? Neah, that's too easy. How about by being calm for a change? >brad@looking.on.ca (Brad Templeton) writes: >>It is not efficient in human or machine resources for a discussion group >>of 100 people to have a discussion distributed to 20,000 machines. In the > >Support this statement - if you can. If we were to set a rough limit based on what Brad thinks is a good idea or what you think is a good idea, you'd lose. Now learn a few things about inter-personal interaction and maybe everyone won't spend all of their time laughing at you. Comp.groupware, indeed. I thought this was going to be an interesting group when I voted for it, not just your private soapbox.
brad@looking.on.ca (Brad Templeton) (06/01/91)
In article <1991May31.192813.28009@gorm.ruc.dk> david@gorm.ruc.dk (David Stodolsky) writes: >>It is not efficient in human or machine resources for a discussion group >>of 100 people to have a discussion distributed to 20,000 machines. In the > >Support this statement - if you can. I hope the question of machine resources doesn't need support. However it ties to human resources. A very large number of sites no longer get all groups. Even sites that get all groups tend to, under human control, treat them differently with different expire times, or different feedings to downstream sites. The more groups a machine gets the shorter the expire time has to be for a given size disk. This is a machine resource that affects humans. If a group has 100 readers and 10,000 sites, then a vast number of sites either have to remove the group manually or reduce the expire time (eventually) on the whole thing, or buy more disks. All are human efforts. Reduce expire times also affect the people reading the groups you want. There are some machines which USENET does not impinge on, but for most of us with limitations on disk and phone time, every new group costs. Of course, I have done something about that, and on my own machine, I am fed exactly the set of current groups being read on my own site and downstream sites, and this is done without human intervention. This is what I belive the long term solution to be. However, until everybody uses a dynamic feeding scheme of some sort, and as long as usenet grows to fill the disk space available, it will continue to be inefficient in the extreme to have a group of 100 people distributed to 10,000 or 20,000 machines. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
emv@msen.com (Ed Vielmetti) (06/01/91)
In article <1991Jun01.041929.8253@looking.on.ca> brad@looking.on.ca (Brad Templeton) writes:
Of course, I have done something about that, and on my own machine, I am
fed exactly the set of current groups being read on my own site and
downstream sites, and this is done without human intervention. This is what
I belive the long term solution to be.
How do you discover new groups? What if someone says "hm, i wonder
what is in alt.industrial, has anyone said anything interesting in
there in the last month?" You can't believe that a 45 character blurb
approved by someone who doesn't even read the group is going to help
you there. It's mighty handy to keep around a day or two of news in
even the most random and banal of newsgroups on the off chance that
something interesting will pop up.
this is especially true of news reading schemes which cut across a
broad swath of groups looking for interesting things. many topics
don't have their own newsgroup, and there's no good guess where the
next bit of interesting materials will pop up.
However, until everybody uses a dynamic feeding scheme of some sort, and as
long as usenet grows to fill the disk space available, it will continue to
be inefficient in the extreme to have a group of 100 people distributed to
10,000 or 20,000 machines.
groups of 100 people don't tend to have newsgroups. remember the
lurker factor, it's a considerable number of folks who never post
anything but still read the net. (hi lurkers!) inefficient netnews
systems tend to get squashed out pretty ruthlessly, anyways -- if
people want dynamic feeding, i'm sure that once they're convinced of
its utilities and cost savings they'll start to use it.
in addition, your licence agreement for dynafeed puts it out of the
reach of a number of systems which might otherwise consider it.
--Edbrad@looking.on.ca (Brad Templeton) (06/02/91)
In article <EMV.91Jun1010753@bronte.aa.ox.com> emv@msen.com (Ed Vielmetti) writes: >How do you discover new groups? What if someone says "hm, i wonder >what is in alt.industrial, has anyone said anything interesting in >there in the last month?" You can't believe that a 45 character blurb >approved by someone who doesn't even read the group is going to help >you there. It's mighty handy to keep around a day or two of news in >even the most random and banal of newsgroups on the off chance that >something interesting will pop up. Mighty handy, but mighty expensive. The way I do it is that an empty spool for every group exists on my machine. Any reader can subscribe to such a group, and they will see nothing there (except perhaps crossposts) However, within a few hours, the last 80 articles (ie. old articles, unless there are 80 new in the few hours) appear in the group, and it keeps coming until the user unsubscribes again. You are limited to this on a uucp system. On an internet system, a better idea would be to have what was eventually planned for NNTP. If a user wants to look at a group you don't get, you talk to an NNTP server that does have the group somewhere out on the net. (This requires the concept of feed servers and "server" servers.) If the user likes what she sees, a feed is set up. (Assuming this is allowed by admins, of course) > >groups of 100 people don't tend to have newsgroups. remember the >lurker factor, it's a considerable number of folks who never post I am aware of this, however, the argument I was refuting was "if 100 people want to read it, it's more efficient as a newsgroup than mailing list." This argument is the old one at the heart of the voting scheme. >in addition, your licence agreement for dynafeed puts it out of the >reach of a number of systems which might otherwise consider it. Perhaps you are unfamiliar with the licence then. Nobody pays for dynafeed. If a site is a professional uucp feeding site like uunet (of which there are perhaps only a few dozen) I ask them to do me a favour (not money) in return. I would not call this putting it out of reach of a number of systems, or unreasonable. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
dmn@cathedral.cerc.wvu.wvnet.edu (Dan Nichols) (06/03/91)
In article <EMV.91Jun1010753@bronte.aa.ox.com>, emv@msen.com (Ed Vielmetti) writes: > > How do you discover new groups? What if someone says "hm, i wonder > what is in alt.industrial, has anyone said anything interesting in > there in the last month?" You can't believe that a 45 character blurb > approved by someone who doesn't even read the group is going to help > you there. It's mighty handy to keep around a day or two of news in > even the most random and banal of newsgroups on the off chance that > something interesting will pop up. > > this is especially true of news reading schemes which cut across a > broad swath of groups looking for interesting things. many topics > don't have their own newsgroup, and there's no good guess where the > next bit of interesting materials will pop up. > I agree. We are doing work on coordination technology for cross- functional teams and there are no specific newsgroups that deal with this area. I have found articles in comp.groupware and even in sci.virtual-worlds that are of interest. Who knows where else articles that are relevant to our work are lurking. > > groups of 100 people don't tend to have newsgroups. remember the > lurker factor, it's a considerable number of folks who never post > anything but still read the net. (hi lurkers!) inefficient netnews > systems tend to get squashed out pretty ruthlessly, anyways -- if > people want dynamic feeding, i'm sure that once they're convinced of > its utilities and cost savings they'll start to use it. > > in addition, your licence agreement for dynafeed puts it out of the > reach of a number of systems which might otherwise consider it. > > --Ed Yes, the members of our group can be classified as "lurkers" right now. In the future we may be posting articles to this group, but for now we are only reading the articles. So I guess the moral of this is that there are lurkers out there who are reading this group and find it quite useful. -------------------------------------------------------------------------- Disclaimer : These are my opinions -- get your own. Dan Nichols (dmn@cerc.wvu.wvnet.edu) Concurrent Engineering Research Center (Making the world safe for CE) 955 Hartman Run Road Morgantown, WV 26505 -------------------------------------------------------------------------
david@gorm.ruc.dk (David Stodolsky) (06/04/91)
brad@looking.on.ca (Brad Templeton) writes: >Instead, groups are defeated by opposition. What an odd system. All current methods for newsgroup creation are defective, because they do not help the interested people select a good name. I do not count "ask a knowledgeable person" as a solution to the name selection problem. I suggest the following: 1. Names suggested during the discussion period are collected by the organizer. 2. Interested persons are asked to rank the names in their preferred order. 3. Single transferable vote counting is used to select the preferred name and determine if there are 100 persons in support of the selected name. This preserves the spirit of the original Guidelines and integrates the results of the preferential voting methods poll of last year. All of the technical details for doing this have been worked out long ago. Multiple name votes have become common, but each time the method is reinvented. Some of these reinventions are incomprehensible and others do not really give the voter a choice. It is time for the Guidelines to *guide* instead of follow what is actually being practiced. The suggested improvement will not do that, but at least we can start to catch up. Renaming will have to be dealt with before the Guidelines will be able to guide. (Parallel votes do not solve anything because, they includes all voting paradoxes and adds some new ones) With every passing day Trial.* becomes less attractive to group champions in comparison with votes, because of the growth of the Net. The newsgroup formation in Alt.* will continue as a monument to Usenets inability to deal with reality :-). If the Guidelines are not updated they should be junked as an irrelevant exercise in bureaucratic malfeasance. -- David S. Stodolsky Messages: + 45 46 75 77 11 x 24 41 Department of Computer Science Tel: + 45 31 95 92 82 Bldg. 20.1, Roskilde University Center Internet: david@ruc.dk Post Box 260, DK-4000 Roskilde, Denmark Fax: + 45 46 75 74 01
brad@looking.on.ca (Brad Templeton) (06/04/91)
There are two things wrong with votes, and particular STV, for picking names. a) The most important thing about a namespace is that it be consistent. Selecting names by popular opinion will pick the most popular name on a case-by-case basis. A different set of voters would thus be picking each name. Does anybody believe this is seriously a way to make a consistent namespace? b) STV, while intellectually an interesting system, confuses a remarkable number of people. Even the reasonably educated Hugo awards constituency (where it is used) is constantly confused about it and needs constant explanation and re-explanation. The same factor has been seen on USENET with STV proposals. We get enough problems with the remarkably simple system we have now, and any controversy results in an eruption of accusations, calls for checks and worse. You think STV would do better? Naming is not an important issue. Because it is the only issue on USENET, it gets blown up way out of proportion. (Also because name and distribution are still linked, unfortunately.) Suffice to say that nobody else ever gives names a second thought for online organization. It's not even considered an issue anywhere else, and on USENET, naming debates generate megabytes a week. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
bob@teamate.UUCP (bob) (06/04/91)
By the end of this week we will have the link completed to map
newsgroups into TEAMate topics (TEAMate is our UNIX BBS product).
This means that all articles will be fully indexed (including text).
AND you will be able to view the indexes, this means that you can
look at the words and frequency counts in a newsgroup to see whats there.
Then do a full content search on the word or whatever.
If anyone would like to give it a try send me email and I'll arrange
access..unfortunately we are only a uucp site so users will have to
get in via public network.
--
MMB Development Corporation, 904 Manhattan Ave, Manhattan Beach, CA 90266
"The TEAMate UNIX BBS for SUN, DEC, IBM, AT&T, HP and INTEL"
VOICE: (213) 318-1322 or bob@teamate.uunet.uu.netdavid@gorm.ruc.dk (David Stodolsky) (06/05/91)
brad@looking.on.ca (Brad Templeton) writes: >There are two things wrong with votes, and particular STV, for picking names. The "votes" are really interest surveys. They are the only accepted method for newsgroup creation for the moment. I would like to address a concrete problem in this thread. >a) The most important thing about a namespace is that it be consistent. Not self evident. Are we not talking about groups of people discussing various topics? The question then becomes a totally different one. >b) STV, while intellectually an interesting system, confuses a remarkable >number of people. Even the reasonably educated Hugo awards constituency If people can not stack up the names in their preferred order they should let their betters select the name. People do not have to understand the system to participate effectively. >Naming is not an important issue. Because it is the only issue on USENET, >it gets blown up way out of proportion. (Also because name and distribution This is why STV is needed. Saying it not important will not stop it from getting blown up way out of proportion. -- David S. Stodolsky Messages: + 45 46 75 77 11 x 24 41 Department of Computer Science Tel: + 45 31 95 92 82 Bldg. 20.1, Roskilde University Center Internet: david@ruc.dk Post Box 260, DK-4000 Roskilde, Denmark Fax: + 45 46 75 74 01
brad@looking.on.ca (Brad Templeton) (06/05/91)
In article <1991Jun4.170128.26382@gorm.ruc.dk> david@gorm.ruc.dk (David Stodolsky) writes: >brad@looking.on.ca (Brad Templeton) writes: > >>a) The most important thing about a namespace is that it be consistent. > >Not self evident. Are we not talking about groups of people discussing >various topics? The question then becomes a totally different one. Quite simply, what is the purpose of a newsgroup name? These are the main ones: a) It identifies the group to outsiders, helps them find it and tells them something about what might be discussed within it. b) To insiders, to a slight degree, it provides a very terse definition of the type of discussion that will go on within the group. c) The hierarchy part affects default propagation at some sites. This is not so much a purpose any more as a consequence. It used to be a purpose, in in some hierarchies, like "talk" it perhaps still is. For (a), consistency is very important. You can't tell one group from another or search through the names if there is no rhyme or reason to the naming system. It is more important that the name fit an understandable system than that it be perfect for (b). The importance of (b) gets overstated. The people and discussion there affect the character more than anything. Many people are unaware of this, as we often get proposals for group names that are nothing but acronyms known only to the initiates. An acronym unknown to outsiders is a bad name. The insiders already know what the group is about. There is little need to shorten the name as you almost never type it. (c) is a problem. This was an original purpose, but the purpose has grown less important and more the subject of argument. Fewer and fewer sites subscribe strictly at the hierarchy level these days, and tools are available to avoid the need for this at all at any feed-cost-conscious site. >If people can not stack up the names in their preferred order they should >let their betters select the name. >People do not have to understand the system to participate effectively. It is not a problem of participation. It is a problem of endless argument. There is already too much argument over the technicalities of the rather simple existing semi-formalism. To complicate the rules invites more chaos. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
chip@tct.com (Chip Salzenberg) (06/05/91)
According to david@gorm.ruc.dk (David Stodolsky): >brad@looking.on.ca (Brad Templeton) writes: >>b) STV, while intellectually an interesting system, confuses a remarkable >>number of people. Even the reasonably educated Hugo awards constituency > >People do not have to understand the system to participate effectively. But if they don't understand the system, they won't shut up when their pet proposals fail. -- Chip Salzenberg at Teltronics/TCT <chip@tct.com>, <uunet!pdn!tct!chip> perl -e 'sub do { print "extinct!\n"; } do do()'
brnstnd@kramden.acf.nyu.edu (Dan Bernstein) (06/12/91)
In article <1991Jun4.170128.26382@gorm.ruc.dk> david@gorm.ruc.dk (David Stodolsky) writes: > >b) STV, while intellectually an interesting system, confuses a remarkable > >number of people. Even the reasonably educated Hugo awards constituency > If people can not stack up the names in their preferred order they should > let their betters select the name. Whose leg are you trying to pull? If you give me a group name I'll give it a rating---a number between -10 and +10, for instance, with -10 meaning I completely disapprove, +10 meaning I completely approve, and 0 meaning I don't care. If you give me a whole bunch of group names I'll give each one a rating. That's easy. STV with, say, five choices is like saying that I have to assign ratings of +10, +5, 0, -5, and -10. What if I approve equally of two different names? STV simply cannot handle this. What if I can't stand any of the names? I want to give them each a -10, but STV makes me pretend that I can put them on a scale from best to worst. What if I want to express subtler levels of approval than the clumsy gradations provided by STV? There's absolutely no way. STV is, I agree, intellectually interesting. It is also theoretically flawed and practically unusable. You want to measure which name people approve of the most? Fine. Give each person a list of all the names. Set up a scale: Y/-/N, +10 to -10, whatever. Add up the ratings that people give. The end result is, by definition, the total voter approval for each name. There's no STV rules to worry about, no interpretations to argue about, no ambiguity in the results. That's approval voting. It works. ---Dan