kaul@ee.eng.ohio-state.edu (Rich Kaul) (05/18/91)
In article <1991May17.170212.5145@looking.on.ca> brad@looking.on.ca (Brad Templeton) writes: | b) Arbitron stats are becoming less reliable. [...] | USENET has increased an order of magnitude while the number of arbitron sites | stayed constant, or even dropped. Not good. Perhaps it's time to admit that arbitron is now a poor indicator of readership entirely? Many sites, and especially large, new sites have gone to using NNTP feed machines and newsreaders on a variety of other machines, both Unix and non_unix, to serve their populations. This lets them add news without heavily affecting the user distribution and forcing everyone to have an account on a single Unix box where news would be read/stored. The old model of "everybody reads on a (connected) Unix system" is quickly falling by the wayside, but arbitron hasn't changed; even a group of Unix boxes connected via NFS is difficult to grab arbitron figures for. The problem of supplying arbitron information when you don't even have access to, much less administrative control over, systems that can read news via NNTP is a real problem. Until the protocol for NNTP offers some arbitron features, don't expect an increase in the reporting percentages. (n.b. NNTP has it's own statistics gathering programs, but they really don't show the number of users, just the number of connections, which groups were read, how often, how many articles, etc. In some ways it's a more accurate gauge of demand than is arbitron.)
reid@pa.dec.com (Brian Reid) (05/20/91)
You are right that arbitron no longer measures what it used to measure, but it does still measure something. I'm just not quite sure what that something is. When the NNTP spec was first circulated 6 years ago, I pointed out that this protocol would make reader measurement impossible and asked them to change it. They ignored this request, and the current NNTP spec makes it virtually impossible to measure readership accurately. I guess they didn't share my belief that readership was important. Stan Barber and Brian Kantor tell me that the new NNTP spec will have in it a protocol that will permit readership measurement. I would be delighted to retire the 7-year-old arbitron technology in favor of this more modern technique just as soon as it becomes workable. About Spaf and the rules: it is definitely past time to apoint a new King of USENET. It was past time 5 years ago, but not very many people noticed. Spaf became King more or less because he announced that he was King. I think that if enough people offer their fealty to a new USENET King, he or she could be duly crowned. Brian
a3@rivm39.rivm.nl (Adri Verhoef) (05/27/91)
>| b) Arbitron stats are becoming less reliable. [...] >Perhaps it's time to admit that arbitron is now a poor indicator of >readership entirely? Many sites, and especially large, new sites have >gone to using NNTP feed machines and newsreaders on a variety of other >machines, both Unix and non_unix, to serve their populations. This >lets them add news without heavily affecting the user distribution and >forcing everyone to have an account on a single Unix box where news >would be read/stored. The old model of "everybody reads on a >(connected) Unix system" is quickly falling by the wayside, but >arbitron hasn't changed; even a group of Unix boxes connected via NFS >is difficult to grab arbitron figures for. The problem of supplying >arbitron information when you don't even have access to, much less >administrative control over, systems that can read news via NNTP is a >real problem. Until the protocol for NNTP offers some arbitron >features, don't expect an increase in the reporting percentages. I was aware of this. That's why I tried to convince other boxes' administrators to gather statistics for their box and have it sent to a central account: All systems that read news inside my domain should produce arbitron statistics and forward the results to the central account, once per month. After a five day waiting period, the results from the central account are gathered together and the final computing step results in a Netreaders count per newsgroup, which is identical to the arbitron input format. So my domain hides several hosts, of which you'll see only one arbitron report. I know that not every host reports, but you'll get the idea.