reid@Glacier.ARPA (Brian Reid) (02/28/85)
I have been trying to find a way to say this in a logical and persuasive
manner for about 2 weeks now; I am unable to make it either logical or
persuasive. I nevertheless beg for your attention because I believe that the
thought that I am trying to figure out how to explain is important. I am not
one of the Usenet ruling council, but I have been a heavy user of computer
mailing lists and bulletin boards and netnews schemes longer than some of
them have been alive; what I am trying to do is distill my experience in 20
years of using computer message systems into something concrete enough that
it can be discussed.
The proposition that is currently being discussed is the proposition that a
proliferation of newsgroups is bad. There are some valid arguments, based on
the difficulty of deciding what group to post a message in, that support
that view, but most of the recent arguments that I have seen are based on
the "too much traffic" claim and not the "too many decisions to make" claim.
First let me offer some substantive evidence that small special-purpose
newsgroups are not an overwhelming source of traffic. Here is the most recent
set of Rick Adams' numbers that has reached Glacier; I have placed an * in the
leftmost column of the line for each newsgroup that I judge to be
"special-purpose":
No. of $ Cost % of Cumulative
Rank Kbytes Articles per Site Total % of Total Group (Articles/contributor)
1 752.1 73 23.50 9.0% 9.0% net.sources (1.6)
2 543.9 321 17.00 6.5% 15.5% net.politics (3.1)
* 3 377.5 202 11.80 4.5% 20.1% net.politics.theory (4.5)
4 372.2 257 11.63 4.5% 24.5% net.flame (1.8)
5 354.5 178 11.08 4.3% 28.8% net.religion (2.9)
6 301.9 29 9.44 3.6% 32.4% net.sources.mac (2.2)
* 7 222.7 118 6.96 2.7% 35.1% net.religion.christian (2.9)
8 218.5 309 6.83 2.6% 37.7% net.jokes (1.6)
9 193.6 350 6.05 2.3% 40.0% net.sf-lovers (2.8)
* 10 184.8 133 5.78 2.2% 42.2% net.religion.jewish (2.8)
11 178.5 235 5.58 2.1% 44.4% net.lang.c (2.5)
12 166.9 143 5.22 2.0% 46.4% net.singles (1.6)
13 163.6 43 5.11 2.0% 48.3% net.origins (1.8)
14 152.3 152 4.76 1.8% 50.2% net.movies (1.7)
15 151.7 159 4.74 1.8% 52.0% net.auto (1.7)
16 151.3 212 4.73 1.8% 53.8% net.unix-wizards (1.9)
17 148.2 115 4.63 1.8% 55.6% net.women (1.7)
18 141.6 208 4.42 1.7% 57.3% net.music (2.3)
19 125.4 146 3.92 1.5% 58.8% net.micro (2.2)
20 117.3 140 3.67 1.4% 60.2% net.audio (1.8)
21 106.8 95 3.34 1.3% 61.5% net.micro.mac (1.6)
22 104.8 76 3.27 1.3% 62.7% net.abortion (1.8)
* 23 103.6 81 3.24 1.2% 64.0% net.nlang.india (1.9)
24 100.0 116 3.13 1.2% 65.2% net.unix (1.9)
25 98.6 29 3.08 1.2% 66.4% net.philosophy (1.3)
Basically what these numbers say is that 2/3 of the total net traffic is
used up by 25 groups, and that less than 10% of the total net traffic is
used up by groups that Brian Reid judges to be "special purpose". Turning it
around, this means that about a third of the traffic is used up by the
"minor" groups. I personally do not find that to be a burdensome load,
though I am sure that some sites might. It is not the dominant factor in
network costs.
So let us consider what a "minor" group is, and look at the lifecycle of a
new group. Basically what happens when somebody starts a new group is that
most people don't care, and unsubscribe as soon as the first message
arrives. Some small number of people read the group. If the topic "clicks",
then the readership slowly grows, both as new netnews readers join the net
and as unsophisticated readers hear about the existence of the group and are
taught how to subscribe to it. Sometimes these groups die slowly; sometimes
they continue on at a low level. Sometimes (net.politics.theory) they go
completely wild.
By contrast, when a mailing list is started up, it grows much more slowly.
It is a big pain to unsubscribe from a mailing list, so people are reluctant
to subscribe in the first place. The information about the very existence of
the mailing list is not distributed over any regular channel, and so people
do not have any mechanism for learning that there are mailing lists of
interest to them. Essentially the only way to learn about a mailing
list is to be told about it by a person who is already on it, or by a person
who knows about it. Yes I know that the list of lists is circulated from
time to time, but nobody reads it. I am not quite sure why nobody reads it.
I would like to claim that the "essence of usenet", the property that makes
it a different communications medium from anything that has ever existed
before (for better or for worse) is that it is simultaneously topical and
reader-selected. Mailing lists are owner-selected or writer-selected. Usenet
groups are reader-selected. I am at a loss to characterize the essential
psychological difference between them, but I have been an ARPAnet user for
14 years and a USENET user for almost 4 years, and there is a fundamental
psychological difference between the two. Usenet is less intrusive on my
life, it is more controlled by me the reader than by the writer, and it
seems to be more adaptible to change than ARPAnet mailing list schemes.
From where I sit (at the moment a red Balans chair) it seems that the
"perfection" property of a netnews group is that it have a high ratio of
readers to writers, a steady but low article count,and enough writers to
keep people from wanting to nuke it. A newsgroup becomes objectionable when
its reader-to-writer ratio approaches (or even falls below) 1.0. By
contrast, an ARPAnet-style mailing list seems to keep people the happiest
when its information content is lowest. I read about 20 newsgroups; I am on
about 50 mailing lists. It would be remarkably nice if all of the contents
of most of those mailing lists could be converted into some form of
newsgroup.
In my opinion this continues to argue for better technology. I realize that
there are sites around the network still running "notes", and running
creaking old versions of netnews, and unwilling to change; nevertheless, if
Usenet is going to continue to be an exciting experiment in communication
technology and not become calcified in a middle-age rut, I claim that
continuing technological advances are worthwhile. Besides Stargate, which is
a wonderful experiment, we should be looking at various kinds of readership
measurement, flow control (only send groups where they are being read),
source quench (dynamically prune distribution trees back to nodes from which
there are distributions being read), interactive probe ("trial subscription"
schemes), and so forth.
One possible way of achieving this is to set up parallel universes. So far
we have had this notion that there is one single Usenet, and that everybody
who is connected to it is connected to the same Usenet. This is a very
democratic notion; has anyone explored the sociological and technological
consequences of doing it another way? What would happen if a bunch of us,
maybe no more than 10 sites initially, were to set up a "C news" or a "D
news" or a "Supernews" network, using new and incompatible software that
tried to accomplish some load management or that tried to have different
editorial policies? Of course that is terribly elitist, but anybody could
join if they ran the new software. There must be other ideas. The thing that
is going wrong with Usenet is not that it is dying in its own weight or that
it is being used by a new class of infidels who don't have the true faith.
The thing that is going wrong with Usenet is that it is going technologically
stale; it is becoming a service instead of an adventure. There's nothing
wrong with a service--I, for one, would continue to read it--but there is no
reason why we have to limit ourselves to that.
--
Brian Reid decwrl!glacier!reid
Stanford reid@SU-Glacier.ARPAkarl@osu-eddie.UUCP (Karl Kleinpaste) (03/04/85)
The claim is made that mailing lists are not a good substitute for newsgroups. In large part, I don't even argue the point. I personally have no objection to a large number of newsgroups, as they are no particular burden on me personally. However, I have to object to some of the criteria suggested for evaluating good points and bad points about mailing lists and newsgroups. ---------- > By contrast, when a mailing list is started up, it grows much more slowly. > It is a big pain to unsubscribe from a mailing list, so people are reluctant > to subscribe in the first place. The information about the very existence of > the mailing list is not distributed over any regular channel, and so people > do not have any mechanism for learning that there are mailing lists of > interest to them. Essentially the only way to learn about a mailing > list is to be told about it by a person who is already on it, or by a person > who knows about it. Yes I know that the list of lists is circulated from > time to time, but nobody reads it. I am not quite sure why nobody reads it. ---------- As moderator for a newly-formed mailing list (mail.firearms, as it's now known), I can speak from very recent experience that these complaints are quite groundless. Several points come to mind: (1) Mail.firearms has grown to some 35 people in less than a month. Technically, this is slower than growth in some newsgroups, as the readership is necessarily limited to those 35 people; but I have already distributed 2 group-originated digests. A third will go out Monday or Tuesday. The digests only started being posted 2.5 weeks ago. (2) The pain of subscribing or unsubscribing is really very little; all one need do is drop me a one-line note in the mail, and *presto* you're {on,off,pick one} the list. No sweat, if you ask me. I must admit that I managed to lose one such request for almost a week, but that was an isolated incident due to my inexperience and lack of organization about the task. (3) The claim is made that information about mailing lists is not distributed over any "regular channel." This is quite incorrect. The author of the above comments even acknowledges that a monthly list-of-mailing-lists is distributed, but then claims that nobody reads it. This is just plain wrong. I have read it every single month since it has been distributed. I have sent a one-paragraph blurb on mail.firearms to Chuq von Rospach for inclusion in his next posting, so I know that this new list will be known there as well. I don't remember just now in what newsgroup the list appears, but wherever it is, I'm subscribed to that group. I believe it's mod.newslists, and that's a "regular channel" if I ever saw one. ---------- > I would like to claim that the "essence of usenet", the property that makes > it a different communications medium from anything that has ever existed > before (for better or for worse) is that it is simultaneously topical and > reader-selected. Mailing lists are owner-selected or writer-selected. Usenet > groups are reader-selected. I am at a loss to characterize the essential > psychological difference between them, but I have been an ARPAnet user for > 14 years and a USENET user for almost 4 years, and there is a fundamental > psychological difference between the two. Usenet is less intrusive on my > life, it is more controlled by me the reader than by the writer, and it > seems to be more adaptible to change than ARPAnet mailing list schemes. ---------- I reject the notion that mailing lists are either owner-selected or writer-selected. I moderate one list and participate in another; neither is so constrained in usage. People wishing simply to *read* the lists are *perfectly welcome*. Again, all they need do is indicate *interest* by mailing a trivial note that anyone can bang out in about 30 seconds. Further, with respect to adaptability, the other list in which I participate (mail.christian, also known as MailJC) has gone through very substantive change in its short life. It started out as a very small collection of people with absolutely no guidelines and less organization. It has since changed character (due in large part to its growing size) to being moderated (thanx to Liz Allen) and having a set of guidelines which keep it running very, very well, so well in fact that I plagiarized them for the start-up guidelines for mail.firearms. I know of no one with objections to its operation, but yet its character has changed to make it progressively better for those interested in it. In starting mail.firearms, there was one individual with questions about the use of a moderator and guidelines; I gave this person the best answers I could as to why such a method was being used, and I have heard nothing more on the subject since. I can only assume that the person has agreed that this is a good way to go. ---------- > From where I sit (at the moment a red Balans chair) it seems that the > "perfection" property of a netnews group is that it have a high ratio of > readers to writers, a steady but low article count,and enough writers to > keep people from wanting to nuke it. A newsgroup becomes objectionable when > its reader-to-writer ratio approaches (or even falls below) 1.0. By > contrast, an ARPAnet-style mailing list seems to keep people the happiest > when its information content is lowest. ---------- So far, mail.firearms has had about 12 people contributing to it, and a few of those were responses to others' postings in the 1st or 2nd digest. The other two dozen or so people are, presumably, reading it for the fun of it, much as people read nesgroups. The information content is very high, that is, it's all very interesting, technical information which is being distributed. I look forward to receiving new contributions, because (so far, at least) they've all been very good. The article count is rather high, I think, since I am about to distribute the third digest of articles. (The first digest had 5 articles, the second had 8, the third will have about 5 or 6 again.) From my experience with this very new mailing list, I have to reject this claim that mailing lists don't have the "perfection property" of a newsgroup. -- Karl Kleinpaste @ Bell Labs, Columbus 614/860-5107 +==-> cbrma!kk @ Ohio State University 614/422-0915 osu-eddie!karl
mark@cbosgd.UUCP (Mark Horton) (03/06/85)
I have to agree with Brian and disagree with Karl on this one. I've also used lots of mailing lists and newsgroups over many years. My experience is that, while mailing list maintainers try to be good about it (as clearly Karl does) things don't always work out so well in practice. To subscribe to a newsgroup, you edit your .newsrc (or if you subscribe to everything, you automatically get it and do nothing.) This takes perhaps 2 minutes. You immediately have access to 2 weeks of back issues. To subscribe to a mailing list, you send mail to the list owner. This first implies knowing about the list and who the list owner is, which is nontrivial. Then you send mail to the list owner. If the mail gets there OK (this implies you found a working route in the UUCP world) and the list owner is responsive, you may be on the list within a day or so. However, it's not uncommon for the request to be dropped on the floor by the mail system or the list owner, or for the list owner not to know how to get mail back to you, or for a delay of a week or more to occur. Then you typically don't have access to back issues (unless you're on the ARPANET and have somehow found out about a place you can FTP them from.) Mailing lists also intrude on your time more than netnews. You get that "you have new mail" announcement and have no idea if it's something important from a coworker or the latest issue of sf-lovers. Once you go into mail to find out, you generally figure you might as well read it. Netnews comes in via a different channel than mail, so you tend to only start it up when you have a little time on your hands. (There are those on the ARPANET who dismis this argument, pointing out that they can have their mass-mailings sent to a different login name, and that they have some wonderful mail interface that understands how to reply to digests and do different sorts of replies to the author or to the group. To them I must point out that very few people have the power to create alternate mailboxes for themselves, and many operating systems, such as System V, just can't do it. In practice, almost nobody uses such features.) Finally, when a mailing list breaks, because a machine or link is down, or a person changes their address, the error message (complete with a copy of the original posting) was designed for the case where the mailing list is small and it's important that the posting get to everyone, so it returns the error to the sender. This person usually can't do anything about it, so he just throws the message away. There are a few mailing lists that have provisions for returning the errors to the owner of the list, but the way to do this isn't universally agreed upon, and most mailing lists don't do it. I know that if I send a piece of mail to any large mailing list (such as namedroppers) I'll get back 3 to 5 warnings from some mailer that it's been a day and my mail hasn't gotten through to somebody at some machine that's down, and a few days later when some of these machines are still down, I get another copy telling me it's been thrown away. We use mailing lists a lot, but for different sorts of things than public discussions ala netnews. We use them for internal things within my project at Bell Labs, for example, we have a mailing list for everyone in my group, and one for everyone in the department. There are several mailing lists used for the UUCP project. I also belong to a few ARPA mailing lists that aren't gatewayed into Usenet. However, traffic on these lists is pretty light, and most of them are intended for private discussions and announcements. If you're conducting any sort of a public discussion where anyone can join in, I prefer to use a newsgroup. Mark
chuqui@nsc.UUCP (Chuq Von Rospach) (03/07/85)
In article <926@cbosgd.UUCP> mark@cbosgd.UUCP (Mark Horton) writes: >I have to agree with Brian and disagree with Karl on this one. I've also >used lots of mailing lists and newsgroups over many years. My experience >is that, while mailing list maintainers try to be good about it (as clearly >Karl does) things don't always work out so well in practice. From my experience, mailing lists are advantageous only in limited areas: a small and specialized area of knowledge where the number of people who want the information is limited; a topic of information where response time is important; a topic that is going to have a known limited lifetime; and a topic where you really don't want to get sidetracked by a lot of comments from 'the peanut gallery' (no offense intended to the peanut gallery). For example, unix-wizards would make a terrible usenet mailing list. Everone wants it, the data on it isn't time critical, and it will go on forever. Singles would make a terrible mailing list for the same reasons, as would most of the other groups currently on usenet. But there are places where I've used groups to great advantage, and where they seem to work well. mail.feminists works-- when net.women.only didn't work as hoped, they moved it to a mailing list. This gives them control of distribution, a moderator to keep the garbage out, and keeps it focused on the areas that need to be focused. I myself have run two major mailing lists in the last year. The first was an unqualified success and was a group of people that collaborated on rewriting the emily post document that you all read in net.announce.newusers (you HAVE read it, haven't you????). It was set up as a mailing list for a number of reasons: . limited audience-- as we put together the first draft, I only wanted to handle comments from people who were actively interested in working on the document. . time critical-- If I had to wait two to three weeks for net propogation on drafts and comments, I'd STILL be working on it. turnaround was on the order of 36-48 hours. . low signal/noise ratio-- If I had made net.news.emily for the development, everyone would have thrown in their two cents worth, and we would have spent enormous amounts of energy arguing about nitpicky items. By keeping it to a private list and posting checkpointed drafts to the net, everyone got their say, but we didn't get bogged down in the trivia. Emily came together quickly and worked well (it still works well after a year, although I'm thinking of doing some minor revisions) . it went away. When emily was finished, the mailing list faded into the sunset. Try that with a newsgroup. The latest mailing list, lan-news, has faded into disuse without really accomplishing anything. I think moving it into a mailing list may have been detrimental over a newsgroup because it made it easier to lose momentum-- it probably ought to have had wider access than it did. Generally, whether or not to go to a mailing list depends on the information-- somethings seem to fit that form of communication well, others don't. We ought to keep both options open, which is why I'm keeping a list of available mailing lists. chuq -- Chuq Von Rospach, National Semiconductor {cbosgd,fortune,hplabs,ihnp4,seismo}!nsc!chuqui nsc!chuqui@decwrl.ARPA Be seeing you!