laura (06/29/82)
Sending netnews is not a right which every good north american (oops, austrailian too) has written indelibly upon their sacred constitution and unix license. Netnews is not free. If it were, one could argue that anyone has the right to post anything they want in a news group like "net.test". The last time I checked, though, there was a substantial phone bill which our site had to pay. I have done a quick-and-somewhat-dirty check of what net.test is costing this site, and now am ready to join up on the side of the mush heads. Calls from utzoo (Toronto) to decvax (Merrimac) made after 1200 midnight come at the relativley cheap rate of $0.25 a minute. I just sent a test article to "to.hcr" a group which exists on our site and on hcr, a site that connects to us. It had a simple text "this is test of the cost of net.test". It took 17 seconds (timed from the last digit dialed to the autodialer returning to normal state). I dont know how much of uucp is overhead, but, chop off 2 seconds as overhead (which will make the math easier). 15 seconds is 1/4th of a minute. that means, had that call come from decvax, it would have cost us 8.25 cents. At three net.test articles a night (the mean over the last 2 weeks) that is 24.75 cents. At 30 days a month that is 742.50 cents. This is *very expensive* on a site which up until recently did not have net.space or fa.polisci or fa.armsd on the grounds that *it was too expensive*. Even if I have grossly underestimated the uucp overhead time, if 50% of the cost is uucp overhead, then I still must pay more than $3 a month for the privledge of seeing that site A now talks to site B when, in more than one case, site A is in the *same room* as site B, and in many cases is a local phone call to site B. We have new sites and new polling frequencies and tests in Toronto all the time -- but nobody has to pay for them. On the whole, i cant think of a single reason for net.test's existance -- is there any reason why the *whole net* should see any test message at all? laura creighton decvax!utzoo!laura
mark (06/29/82)
The official policy on test messages is as follows: A site needing a test should feel free to post a test message to an appropriate test newsgroup. They should use the smallest scope test newsgroup that fits their needs. Usually this means "test", which is local to a particular machine. If network distribution needs to be checked, you can use to.foovax (where foovax is one of your immediate neighbors) or, if that won't work, xx.test, where xx is the smallest containing newgroup class. (For example, here we have newsgroup classes local, osg, cb, btl, bell, and net, in increasing size. I usually use osg for testing.) Only if there is nothing smaller, and if a to.something newsgroup won't do, or if you REALLY NEED to know if your article is getting to the WHOLE NET, should you post to net.test. In the previous message, I also said that the body of the message should say why the test is being conducted. I have since discovered that the first n-1 tests don't work, and you get sick of typing in the same message over and over, so you type "asjfklfd" as a body. When it finally does work (and you are shocked and surprised) the message has already gone out. I recommend that you put the body in a file, and run inews -t test of new version -n net.test < file for the tests - with csh or the history shell, it's easy to repeat these commands. When you do this, please make the subject also somehow meaningful. If there are 2 test messages each night, out of 100 total articles, and if the articles are batched in groups of, say, 15, then the real costs (making and closing down the phone call) are shared over the batch. Also, your phone cost is rounded up to the next higher minute. So if the 1 test message in a batch takes 5 seconds to transmit (it really probably takes less, unless your machine is heavily loaded) it will increase the phone charge by one minute 1 out of every 20 calls. Surely this is insignificant. On the other hand, if decvax calls you on demand, causing a separate phone call for every message, you'll get 9 zillion 1 minute phone calls. This can add up. Finally, each user and each site can easily unsubscribe to net.test. If your site doesn't want it, tell your neighbor not to forward it. If you, as a user, don't want to see it, just unsubscribe. Now, with all these guidelines in existence, let me point out that test messages are essential, and that if we didn't have net.test, people would post their test messages to net.general. We all know how horrible that would be! Mark
mkg (06/29/82)
I agree with Laura. *.test should not be distributed. I have a simple
fix to ifuncs.c that prevents any test message from being transmitted.
The article IS posted on the local system so it can be read by the
submitter to verify that all is well.
diff 2.8generic/src/ifuncs.c 2.8src/src/ifuncs.c
46c46
< if (ngmatch(h.nbuf, srec.s_nbuf)) {
---
> if (ngmatch(h.nbuf, srec.s_nbuf) && !ngmatch(h.nbuf, "all.test,")) {
If you really MUST exercise a link, use to.system!!!
Marsh Gosnell BTL Piscataway (201) 981-2758 npois!pyuxbb!mkgmkg (06/29/82)
I had a discussion with Mark about the philosophy of *.test. I agree with Mark that a good way to handle the "test" problem is to request that your neighbor not send *.test articles to you. Besides, compared to some "active" newsgroups (net.movies for one) net.test is small (miniscule?) potatoes. Marsh Gosnell BTL Piscataway (201) 981-2758 npois!pyuxbb!mkg
mark (06/29/82)
Please do not install March Gosnell's fix. If there is even any HINT that "all.test" behaves any differently than some other newsgroup, people will start posting their test messages to some newsgroup that they are sure IS handled exactly the same way, like net.general. My experience has been that there is no way to keep people from posting test messages, and I'd much rather have them flagged by being in a separate newsgroup so that I, as a user, can turn them off. We have it good in USENET - on the ARPANET, where there are only mailing lists, there is no "test" mailing list (who would be on it? what would it test?) so people mail test messages to real mailing lists. I do agree that there have probably been entirely too many test messages posted lately. I don't know the specifics of any of them, since they all have cutsie bodies, but I would be willing to bet that most of them could have gone on either "test" or "to.xxxx" just as well as "net.test". I'm thinking of making some recording files to make people understand what certain newsgroups are for, such as net.general, net.test, net.jokes. These are probably the most abused newsgroups. A typical recording would read: net.jokes: This newsgroup is for the posting of jokes. If you want to complain that you don't like the joke you just read, please send mail to the author. If you want to discuss jokes in general or guidelines for using this newsgroup, please use net.jokes.d. If there is any possibility that your joke might offend someone, please use net.jokes.q. Anyone have any other newsgroups that they feel get abused a lot? I'm making a whole batch. Mark
felix (06/30/82)
The only reason I can think of having a net.test newsgroup is when somebody having troubles send articles out to the net but everything works perfectly locally, just like the situation I encountered recently. Felix Luk