Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/17/84 chuqui version 1.7 9/23/84; site nsc.UUCP Path: utzoo!utcs!lsuc!pesnta!amd!nsc!chuqui From: chuqui@nsc.UUCP (The Phantom) Newsgroups: net.news,net.followup Subject: net overload-- comments and suggestions Message-ID: <2404@nsc.UUCP> Date: Tue, 26-Feb-85 23:43:45 EST Article-I.D.: nsc.2404 Posted: Tue Feb 26 23:43:45 1985 Date-Received: Wed, 27-Feb-85 10:54:27 EST Distribution: net Organization: The Paradise Lines: 223 Xref: utcs net.news:2763 net.followup:4424 It is becomming clearer and clearer that the network is approaching overload-- we are simply getting too much volume for the average (or even me!) to deal with appropriately. I've been thinking about the philosophical and humanistic approaches to viewing the network, and I've made some basic (and potentially incorrect) conclusions based on my experience. These went out to a good number of the other people who work widely with the network for comment, but I see today comments in net.unix and net.unix-wizards from Mike Muuss and from Lauren about overloading in the arpa/usenet interfaces in those groups, I've seen serious problems in a large number of other groups as well-- net.general, net.singles, just about everywhere. I no longer feel comfortable waiting for a first round of comments before taking my thoughts public-- I'm not at all sure we really can afford to do a lot of arguing-- we've done a lot of that without really getting anywhere. It's time to do something about the geometrically increasing volume we're seeing, ESPECIALLY the volume of unneeded duplicates and useless garbage showing up. The following are my thoughts on the subject. Many are controversial, some may be unworkable, and all of them are guaranteed to upset someone. I think we need to realize the net is too large to keep EVERYONE happy, and start realizing that tradeoffs need to be made. if we don't do them in a coordinated and thought out manner, I'm afraid someone will force it on us-- Arpa groups will cut off and limit our access to their knowledge, backbones will arbitrarily cut out groups for cost reduction, or certain groups and users we can ill afford to lose will simply get tired of the junk and leave the net. It's time to do something, folks. I suggest heartily that comments you have on this be made to me by mail unless you really have something the net needs to hear. I will be happy to summarize the feedback I get and modify my proposals as needed, but I want to minimize the added amount of unneeded volume we have going around. chuq ---- [Ed note: this is excerpted from private mail, the starting point being that the current usenet documentation does more to confuse the user than inform them, and goes off onto my normal tangents...] I agree with Mark and Roy-- the documentation is woefully inadequate. I think the biggest reason for that is the fact that the audience has changed, and the net hasn't kept up with it. I don't agree with the thinking that a technological answer can be found. At the basest level, all we'll do is start up the fascist wars, and all the SA's that won't be bothered will simply refuse to upgrade or comment out the 'offending' code. The places where the code is needed most won't have it anyway, and those that do won't really need it-- we'll be inhibiting the parts of the user community that don't need inhibition. Realistically, if we can't implement a software solution at the backbone sites, we can't expect it to work-- aliases seems to work well for mapping bad groups, because things get fixed up as the news moves around, but controlling users can only be done at that level if we set up 'hit' lists, something I think I'm philosophically opposed to. The net has changed, but I don't think our attitudes have changed as fast. We still look for technological fixes for problems-- this might have worked with 50 or 100 sites that cooperated with the net, but with 2000 sites running A news, B news (18 versions), notes, mlisp in emacs, and GLOS knows what else, getting a global solution installed is useless. We can fix parts of the problems at the backbone level, we can fix parts of the problems by fixing news (and notes) and cajoling people to upgrade, but we can only do so if people think it is in their own interest. SA's won't consider anything that takes part of their time to be in their interest, no matter how small the time, no matter how large the benefit, simply because they don't want to have to think about netnews and maintenance. This is similar to the problem of message responsibility. If we required SA's to monitor their users postings, or to sign a note admitting responsibility for anything, they simply wouldn't be on the net. Officially, at least. Perhaps it is time to stop thinking about usenet as a collection of hardware and software-- those have become quite secondary to the people involved-- and look for humanistic solutions to our problems. The first step is documentation. Perhaps it is time to do to our introductory documentation what I did to Emily Post a while back (was that a whole year ago?? sigh). I might even volunteer to rewrite the silly thing, if nobody else will. It should emphasize how to work with the network, not how the network works-- oriented towards the people and not the technology. The other thing I think we need to look at seriously is the user interface to the news. Not just the program-- rn is a BIG step forward, and I'd love to see readnews and vnews bite the dust-- but the actual structure of the network, especially in the topic naming space. There are some large and serious problems: o Misnamed groups. Many groups are incorrectly or ambiguously named. o Too many groups. Too many groups have been created for the wrong reasons, and simply don't go away. o Static and inconsistent naming spaces. The naming space doesn't change as the need changes, it simply expands. The expansion, even worse, is haphazard. There is no way for a beginning user to even start to figure out what is happening-- he has no guides. I'd like to mention the concept of 'chunking' here. Essentially, psychologists have found that people can put together groups of about 7 to 10 objects and understand them. A good naming space would allow us to 'chunk' out the various topics in ways that allow people to comprehend the boundary levels. There are a lot of things that we can do to smooth out the network interfaces with little effort, if we decide to do them. First, we need to trim the deadwood (back to 'nuke wobegon', right chuq? right....). I've fought this battle before, and it might be time to fight it again. The basic argument against getting rid of groups is technological (the software can handle it, what bother is it?). The problem, however, is humanistic-- I don't care WHAT the software can do, if the people can't deal with it,, and I think the net is showing signs of people's (especially novices) ability to cope with the complexity of the naming space. The first level of reducing complexity is removing those groups that aren't being used-- nobody will really notice, but the novice won't look at an active list and wonder if maybe the message ought to go there instead (or worse, too!). The second step ought to be recognizing the advantage of general purpose topics. Why create net.psych when net.sci allows a greater range of related topics to share the name space? A good example of this is the set {net.lan,net.dcom,net.lsi}, all of which could be compacted into net.hardware with no loss of usability with the advantage that all hardware now has a home. We can make similar comments about other groups-- net.research, net.usoft, net.cog-eng, and some others could all become net.software. Now, instead of 10 groups, we have two, with solid, simple topic definitions that don't overlap with other groups and allow a wide range of topics to be discussed. As an alternative I point to the recent discussions in net.tv.drwho about the inappropriateness of Prisoner discussions in that group-- valid because of the limited appeal of the group. I'm planning on creating net.tv.pbs as an alternative-- general purpose enough to be used for a wide range of discussions. I wish I'd seen this before I created net.tv.drwho, as a matter of fact. But I think that for the naming space to be really useful it needs to be flexible. That means easy construction of topics when they are useful AND easy destruction of them when they stop being useful. Again, this has been proposed, and the main arguement against it is technological-- 'the software can handle it, leave it around in case we need it again'. From a humanistic view, it adds complexity, we ought to remove it and recreate the damned thing if we DO need it again-- it got created once, didn't it? At this point the naming space is a significant disadvantage to us because I and others fight creation of almost ALL groups simply because once they exist they NEVER go away. We don't have groups we could use (such as net.sources.msdos) because we're afraid to create groups we might not be able to get rid of later-- essentially the existence of net.wobegon keeps us from having net.sources.msdos. You figure out whether the existence of an unused group for the listeners an NPR radio show is worth more than the existence of public domain ms-dos software. I have my own ideas, but I also won't allow creation of groups without a precedent of existing volume simply because we don't have a way of removing those without volume. So, we can't get rid of wobegon, and we can't create msdos because of it. There is a third level involved as well. From a humanistic viewpoint, I think certain newsgroups are detrimental to the overall message content of all groups on the network. We say we are against the rabid flaming that is engulfing the network (something I, unfortunately, let myself get into recently as well...) yet we not only tolerate it, we give flaming its own group. That isn't toleration, that is outgright acceptance. The argument for net.flame that 'it keeps the flames off the rest of the net' doesn't really wash. It doesn't, as any group can tell you. What I think it DOES to is semi-conciously tell people that FLAMING IS OK, when it really isn't-- flaming takes a lot of bandwidth, flaming causes a lot of pain, and except for certain personal satisfactions for getting something 'off my chest', it doesn't do a single positive thing for the network except raise phone bills. I'd like to see net.flame abolished, and have it made perfectly clear that people are expected to act in a mature and professional manner-- we COULD make it stick, you know. If people know that they are going to get beaten bloody for flaming, they'll be more careful. But now, a few of us 'self-righteous fascists' scream, but flaming is really acceptable, so nothing much happens. I could make similar arguments for other groups as well-- net.abortion, net.origins, the proposed net.porn-- all were set up to 'isolate' the discussion, but none of the discussions really stay isolated, they insist on moving into other groups because the zealots want you to listen ANYWAY, even if you are unsubscribed. My preference is to get rid of the groups, and if the discussions that do come up get out of line, stomp on them. I also think it's time to reconsider net.jokes, but I'm running long as it is. It's time to look at the net from the point of view of the person-- especially the naive person-- using the net. To make the net effective for them we need to take steps to make it understandable. I think we can do this AND make the net better for the experienced user because the noise level can be seriously reduced. My basic recommendations are: o Write decent humanized documentation. We no longer care how the network works, but how to work on the network, and what the implications of various actions are. o Improve the simplicity of the user interface. Rn is the biggest step in the right direction I've seen yet. o Improve the simplicity of the network. Removing unneccessary groups, rationalizing the naming space, and making the naming space flexible are all strong steps forward in this direction, and better yet, can all be handled administratively. o Make a committement to the quality level of the postings. This means making sure that people realize that certain forms of postings (flames, 80 line inclusions of someone elses article followed by a one line riposte, unfunnny and repetitive and repulsive 'jokes') are not acceptable in a mature and professional atmosphere. I think if we take these steps, we can reduce our current noise level problems, we can make the net more useful for everyone, and we can solve at least some of our technological problems (such as modem bandwidth) as well. comments, of course, are welcome. I realize that some of my suggestions are rather revolutionary, but I think it is time for a revolutionary change in thinking. Usenet is no longer driven by the technology or the software, and hasn't been for a while. It's time we realized that. chuq -- From behind the eight ball: Chuq Von Rospach {cbosgd,fortune,hplabs,ihnp4,seismo}!nsc!chuqui nsc!chuqui@decwrl.ARPA We'll be recording at the Paradise Friday night. Live, on the Death label.