reid@decwrl.dec.com (Brian Reid) (01/07/88)
A couple of months ago Brad Templeton convinced me to adjust the "cost per reader" figures to include propagation. This seemed reasonable: if a newsgroup only has 50% propagation, then it has a cost per kilobyte that is only half of what a 100% group would have. Alas, since I'm a dumdum, I multiplied by the propagation percentage, e.g. 50, and not by the propagation fraction, e.g. 0.50. OK, I admit I should have tested the change, but it seemed so simple. As a result, the numbers for December are a factor of 100 too large. I'm not going to post a revision: even USENET readers can divide by 100. Brian
jerry@oliveb.olivetti.com (Jerry Aguirre) (01/07/88)
I would suggest that the most serious error in the cost/reader computation is the one that has the least comment on its accuracy, the $.10/minute value. I don't know how this figure was arrived at but I can think of at least two reasons why it is too high. The first is that usenet sites tend to be grouped into clusters. Looking at the New Jersey and the California bay maps show hundreds of sites that can call each other with a local call. I don't pay anything for my two main feeds, they are local calls. Two of my "leafs", Tymix and 3comvax are in the stats and I don't think they pay that kind of rate either. (And, of course, phone company sites may or may not pay for calls and that may or may not represent "real" money.) The second reason is that some of the connections are over dedicated links. Of the 5 sites that I post info on 4 of them are connected over ethernet and don't pay any phone charges. In some cases even some of the long-distance connections are over dedicated (fixed cost) links. (I notice that the decwrl-decvax link has a pathalias cost of "LOCAL".) I know that there are a few sites that are paying a lot more I just don't think that they will average out with the number that are paying zero. So the estimated cost is higher than most pay, not an average, and lower than some sites pay. It just isn't any more meaningful than the bytes/reader data. Collecting actual costs on a month to month basis is impractical but couldn't the arbitron configuration include a field to specify whether the feed(s) were of a particular type, say dedicated, local-call, zone, long-distance, long-distance-night, pc-persuit, or arpa? With that kind of data the cost estimate might be a little more believable. Jerry Aguirre Systems Administration Olivetti ATC
nyssa@terminus.UUCP (The Prime Minister) (01/07/88)
In article <228@bacchus.DEC.COM> reid@decwrl.UUCP (Brian Reid) writes: >I'm not going to post a revision: even USENET readers can divide by 100. Is that 100 binary, octal, decimal, hex or other? :-)
page@ulowell.cs.ulowell.edu (Bob Page) (01/08/88)
A significant percentage of the sites that ship news around over great distances (the 'backbone'-- however you define that) use NNTP over the Internet. That would further reduce the 10c/min transmission cost figure that Arbitron uses. ..Bob -- Bob Page, U of Lowell CS Dept. page@ulowell.edu ulowell!page "I've never liked reality all that much, but I haven't found a better solution." --Dave Haynie, Commodore-Amiga
reid@decwrl.dec.com (Brian Reid) (01/09/88)
I used 10 cents/minute for arbitron because that's the same number that Rick Adams used (and uses) for his "Traffic for the last 2 weeks " postings, and I wanted people to be able to compare the numbers for accuracy. (The programs were independently written and work on slices of USENET taken across the country from one another, and any discrepancy was interesting to me). The true cost of carrying a newsgroup is more than just the transfer charges; there is disk space, CPU time used to run the news software, time spent expiring it, and so on. Maybe 10 cents per minute is the wrong metric, and maybe using dollars is the wrong metric, but the *relative* values of the numbers are quite accurate and it is appropriate to have some non-zero cost per kilobyte of news. I think I'll start labeling them not in dollars but in Zlotys or Yuan (two currencies without officlal exchange rates into dollars). Brian
geoff@desint.UUCP (Geoff Kuenning) (01/09/88)
In article <12330@oliveb.olivetti.com> jerry@oliveb.UUCP (Jerry Aguirre) writes: > I would suggest that the most serious error in the cost/reader > computation is the one that has the least comment on its accuracy, the > $.10/minute value. I agree with Jerry that this is an inaccurate value, but my solution is to look at the cost/reader as being the middle of a three-orders-of-magnitude estimate. In other words, a newsgroup listed as $2.00/reader actually costs somewhere between $0.20 and $20.00. Yes it's crude, but then so are the statistics (as Brian is well aware). -- Geoff Kuenning geoff@ITcorp.com {uunet,trwrb}!desint!geoff
michael@stb.UUCP (Michael) (01/16/88)
(Tried to mail this, but it bounced) Rather than Zlotys or Yuans, how about real currency like Zorkmids or Federation Credits or ... -- : Michael Gersten ihnp4!hermix!ucla-an!remsit!stb!michael : sdcrdcf!trwrb!scgvaxd!stb!michael : "A hacker lives forever, but not so his free time"