tale@cs.rpi.edu (David C Lawrence) (12/31/89)
In <10371@pucc.Princeton.EDU> IRWIN@pucc.Princeton.EDU (Irwin Tillman) writes: > I posted that article with a 'princeton' distribution, as it was not > relevant to people outside of this area. I'm receiving notices > (read "hate mail") from around the world telling me that the article > reached them. They're wrong to send you hate mail for it. If they looked at the distribution header they would have seen your intent. > I assume that some site is munging the Distribution header. Wrong assumption. Site princeton itself is most likely directly responsible for that article leaving your area; you probably should have spoken to that news admin first about this. I posted a rather lengthy <#5`G1_@rpi.edu> just over two weeks ago to news.sysadmin explaining the Distribution: conundrum. Unfortunately I didn't save it, or I would just have mailed it to you. Perhaps someone that keeps 15+ days of news on their system will see this and be kind enough to forward it to you. (Which is why I posted this reply now in spite of not attempting to explain it again.) Here are the headers of your article: Path: rpi!brutus.cs.uiuc.edu!wuarchive!texbell!rutgers!njin!princeton!phoenix!pucc!IRWIN From: IRWIN@pucc.Princeton.EDU (Irwin Tillman) Newsgroups: news.config Date: 29 Dec 89 20:11:00 GMT Reply-To: IRWIN@pucc.Princeton.EDU Distribution: princeton Organization: Princeton University, NJ Lines: 12 Disclaimer: Author bears full responsibility for contents of this article Dave -- (setq mail '("tale@cs.rpi.edu" "tale@ai.mit.edu" "tale@rpitsmts.bitnet"))
coolidge@brutus.cs.uiuc.edu (John Coolidge) (01/05/90)
[Distribution _intentionally_ expanded to world (originally Princeton)] IRWIN@pucc.Princeton.EDU (Irwin Tillman) writes: >Regarding articles posted with 'princeton' distribution showing up far away >from princeton: >I wrote: >> I assume that some site is munging the Distribution header. >My assumption was wrong. The Distribution header is being maintained. >I've also sent out a sendsys with a princeton distribution, and received useful >info. >It seems that there are a whole bunch of sites that list 'all' in their sys >files. >For example, the path that my article took to get to David was: >>Path: rpi!brutus.cs.uiuc.edu!wuarchive!texbell!rutgers!njin!\ >> princeton!phoenix!pucc!IRWIN >That makes sense to me, up through rutgers. >In wuarchive's sys file, it says that it accepts 'all'. It also says that it >sends 'all' to brutus. More unexpected. >In brutus' sys file, it says that it accepts 'all'. It also appears to send >'all' to rpi. So RPI gets to read stuff irrelvant stuff from Princeton. >Perusing other unexpected 'sys' files that arrived, I noticed other unexpected >flows of 'princeton'. For example, many sites fed by brutus seem to receive >'all'. Same goes for uudell. Here and there, the same thing seems to happen >at wuarchive, bigtex, and texbell, along with their neighbors. So you >people out at Pitt, Ohio State, Harvard, all over Texas (it's a big state), >and here and there at USC, Pennsylvania, Montana, Wisconsin, U. Illinois >all get the read stuff intended for a small part of New Jersey. >Still, it makes the problem seem pretty clear to me: people need to be much >more careful about using the 'all' pattern in their sys files. The problem is >not with the 'princeton' distribution; it's just that some of the sites >receiving that distribution are passing news to their neighbors using an 'all' >pattern. >What's the solution? If you maintain news, make sure you don't >indiscriminately use the 'all' pattern. If someone feeds you something you >weren't expecting (e.g. the "outer-mongolia" distribution), you may pass it >it on downstream. Here's the problem: we (and wuarchive) explicitly take anything sent to us, useful or not. That's why the all/all in transmissions between us (for those who don't use C News, sys files under C use the format newsgroups/distributions for each site). Actually, I don't quite send out all/all, I send out all/all,!srg,!uiuc (in other words, I respect local distributions, but not global ones. Why? The problem is: what's a valid way to respect distributions? Suppose I'm sending now not to wuarchive but to rpi, who really doesn't want everything. I could say: <newsgroups>/<newsgroups>,inet,ny,rpi, <whatever else>. But if I do that, and some wonderful distribution (say 'edu') gets used somewhere, then they don't get it. Alternatively, I can say <newsgroups>/all,!<unwanted things>. But then I need a canonical list of unwanted things, which is difficult to get. So I compromise and send with distribution all, minus things that I _know_ shouldn't go on --- things local to me. If one particular distribution winds up bothering the people I feed, I then kill it off too. The best use of distribution I can see is: watch what _you_ send out from your own local distributions and kill things that shouldn't get out. Once it's past your bounds, you really can't expect others to stop it. --John -------------------------------------------------------------------------- John L. Coolidge Internet:coolidge@cs.uiuc.edu UUCP:uiucdcs!coolidge Of course I don't speak for the U of I (or anyone else except myself) Copyright 1989 John L. Coolidge. Copying allowed if (and only if) attributed. You may redistribute this article if and only if your recipients may as well. New NNTP connections always available! Send mail if you're interested.