lauren@rand-unix.ARPA (02/25/85)
The problems with traffic volume both in these ARPANET lists, and in the linked Usenet groups, is becoming significant on both sides. Simply blaming the Usenet side doesn't do much good--I've seen just as many useless postings from the ARPA side of the fence. And while the cost (other than time and disk space) of each posting on the ARPA side is essentially nil to the participants, the costs are very high, ultimately, on the Usenet side, where in most cases EVERY message gets sent to EVERY machine, usually by dialup phone line at 1200 bps, and frequently via long distance, not local, calls. I for one am trying to discourage the creation of more specialized Usenet groups for awhile, in the hopes of getting people to instead spend the time to establish coordinated mailing lists that will only involve the people who are actually interested in particular topics. As ARPANET people know, mailing lists, if properly managed, can provide much faster distribution than the current point-to-point netnews system on Usenet. Of course, mailing lists are more trouble to maintain than just blasting a message out all over the world (and ignoring the costs) but when people are trying to get real work done (for example, the various collections of people working on the various phases of the UUCP Project) mailing lists can be far more efficient in both time and money than blanket netnews discussions and distributions. Mailing lists, while preferable to netnews distribution in many cases, can still have problems of volume--as we're seeing on ARPANET now. But I think that mailing lists are still superior to netnews distributions in many cases if properly coordinated and planned. But in any case, am I alone in getting the feeling that we (both on ARPANET and even more on Usenet) have crossed over some sort of volume "threshold"? It's getting almost impossible to deal with the volume of submissions being entered, on whatever topics, from an ever growing crowd of users, most of whom have no idea what has ever been discussed in these lists before. As this problem continues to grow, more and more people will be forced to drop off the lists (as they are doing now with Usenet newsgroups) since they simply won't have TIME to deal with all this material, much of which is not very useful and just represents (in many cases) useless quips or repetitive questions/answers. In my opinion, the models under which both the major ARPANET lists and the Usenet groups were founded are not scaling up well to the growing user population, as almost anyone on these lists/groups must realize by now. For Usenet, some progress can be made by discouraging many new newsgroups and promoting coordinated mailing lists as a step forward. On the ARPANET side, where many lists already exist, the next step isn't so clear. --Lauren--
Conde.osbunorth@XEROX.ARPA (02/26/85)
Lauren, Here's an approach taken by some Xerox mailing lists which may be adapted to your situation. Some lists in digest form will mail out the table of contents only. If the user is interested, he will retrieve the entire contents to his machine. The file copy command is typically embedded in the table of contents to make it easier. In the particular mail system that I am using, there is no equivalent of the Unix netnews command to share news messages. I do not know if this is feasible, but here's how this may be adapted to USENET sites. If you are willing to deal with 1-2 day delays in reading messages: - Each digest mails out its table of contents. A non-digest message sends out the subject line only. - A user uses some program to peruse the table of contents (TOC) If the message is available locally, (for some value of local) the user has the option of reading it. Otherwise, it is simply marked for later retrieval. - During that evening, a program will try to retrieve all files which are marked "interested" but is not already available locally. The messages will be retrieved from a set of hosts which may have them. - The following day, the user may read the messages. As an implementation issue, some kind of universal message id scheme and a database could be used to index into message contents/subject lines. This way, the user could ignore all message which say: "What's the termcap entry for a Trash-80?". It may be possible on ARPA sites, but I do not know if this will even be worth considering for usenet sites that redistribute messages to other sites. The hard part is knowing who has the replicated copies/when one is capable of doing cleanup operations (i.e. zapping files) without causing hardship to others. Some kind of expiration date scheme may work too... Daniel Conde conde.pa@Xerox.ARPA