henry@utzoo.UUCP (Henry Spencer) (01/09/87)
Dave Sherman asked a few questions about the format of our uucp status report, and suggested that I might want to post the answers. Seems reasonable, so here goes. > > 1 spectri poll (oldest 3.0 days) > > mail & news I understand. What is poll? ... We run our calling system in a slightly more complicated way than some other sites. Every hour, we try to call every system for which there is traffic. (Of course, some of these calls are blocked by things like time-of-day restrictions before they get very far.) There are some sites that we want to poll regularly, either because they don't call us or because we want to call them occasionally as a safety precaution against things like autodialer failures. This is done by creating spurious traffic for them: an empty C. file is enough to convince the software that there is traffic for that site. Once the connection is made, uucico looks at the file for work requests, finds none, does none, decides that the file's requests have been completed, and removes the file. That's what a "poll" item is. They are listed in our uucp report only if they are old enough to be suspicious. The reason behind this slightly involved scheme is that invocations of uucico can fail for many reasons, such as the line already being in use. Trying to poll people by just running uucico is thus unreliable. The poll-file scheme makes utzoo try every hour until a successful connection is made, and then stop trying until the next time polling is requested (or real traffic gets queued). > And what is class c? Wups, sorry, for obscure historical reasons some bits of news go out as class c rather than the normal d, and the reporting program didn't know about that. Fixed. > > alice 2 > > 2 what? files transferred? what determines the point at which you > won't show bytes/seconds = speed? Yup, just files transferred. Small transfers -- ones with size or duration below certain thresholds -- are considered to be too small for reliable bytes/second calculation and hence aren't counted for that purpose. If there was nothing but small transfers, the reporter just reports a count of them without trying to estimate speed. > > dciem 286089/3484 = 82.1 (50.41 - 95.22) > > 50.41 and 95.22 are what? Slowest and fastest speeds for > transfers of individual files during this period? > And why do you only show this for some links? Yes, they're the slowest and fastest speeds. They're shown only if there is at least 20% deviation from the average. > > utcs 737590/1992 = 370.3 (287.48 - 420.59) > > Wow. 4800 baud direct line? UofTNet? ... 4800-baud direct line. Some day we will be on the local network, but not just yet. We also have a 4800-baud hardwired line to utcsri, but that one is out of order at the moment and our hardware man is too swamped with more urgent things to investigate right away. -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,decvax,pyramid}!utzoo!henry