bstempleton (02/09/83)
I think the uucp and usenet documentation should provide some better advice to people choosing site names. We seem to get all kinds of people who put strings such as "vax", "70", or even "unix" in their site name. When you are limited to 7 characters, why put such useless information in your site name. When major domains start getting set up, this will not be as important, but any site under the main north-american public network domain will have to have something more meaningful than "vax" in their name. One site I asked about this said they put "unix" in their name to point out to management that it was a unix machine that gave network mail. That may be a justification for a while, but in the end I think it's what your machine is for or who owns it that is important, not what it runs or what kind of hardware it is.
bobr (02/14/83)
yes, and while we're at it, choose MNEMONIC names, please. I shudder at the thought of hundreds of houx5, utcsrgv etc machines on the net. When we're teaching first year students how to program we (at least at U of T) IMPLORE them to choose mnemonic variable names but when it comes to naming our machines we can't do it ourselves. <flame off> Christoph Bobrowski, Dept of Computer Science, Toronto
ld (02/16/83)
Sorry Christoph(er?), mnemonic site names and uucp do not necessarily mix. When administering a multi-node-star installation, uucp insists that three rules be honored in naming said sites: 1) Site names must be unique. 2) Site names must be the same length. 3) Site names must be no more than 7 characters. The first and third rules should be obvious to anyone. The second is a little bit more obscure. Due to a prefix scanner in uucp, if site names are not the same length, mail (and news) can get misdirected. A case in point occurred when we had site connections: hpda2 / \ hpda ----- hpda3 If hpda2 sent files via uucp to hpda3, but it was deferred (put in the spool directory), then hpda connected with hpda2, hpda would take ALL files destined for hpda3 because (within 4 characters) `hpda' matched `hpda3'. If rules 2 or 3 are violated, there MIGHT be no side effect. If there is a side effect, it will drive the uucp administrator to drink, trying to figure out where the files went (try this on a non-source licensed system). Then there is netnews! Remember, the return addresses keep getting longer each time the news is passed to another machine. This should give some impetus to keep site names short (5 characters should do it). Superfluous characters (such as `-') should be avoided (yes, I know some hp sites have `hp-', they do not believe in the same religion that I do). If the return path gets too long, it gets mysteriously truncated. What it boils down to is: What are the site names for? Are they there for the sake of the humans or the sake of the transport mechanisms? There is a netnews site directory that equates site names with entities. Besides, what better mnemonic for Indian Hill than ih? Larry Dwyer ucbvax!hpda!ld
mark (02/16/83)
Re: 1) Site names must be unique. 2) Site names must be the same length. 3) Site names must be no more than 7 characters. Close but no cigar. The real rules are 1) Site names must be unique. 2) Site names must be unique in their 7 characters. There is no requirement for site names to be the same length. There is a bug in many UUCP's that barfs if one site happens to be a prefix of another (we hit this one with cbosg and cbosgd), and it's easy to fix. In /usr/src/cmd/uucp/anlwrk.c, find the routine "getwrk". Near the beginning there is a call to "prefix". Just augment this check to also insist that one file name is exactly 5 chars longer than the other. Note that there are many long names on the net, e.g. microsoft, ubc-vision, genradbolton, rochester, etc. It's just that UUCP truncates all names to 7 chars (presumably to leave room for the "C." and "X1234" on the ends of the filenames). You can have longer names, and mail will use the longer name, so the chopping is supposed to be invisible. Kind of like the old C compilers chopping all names to 8 chars - this didn't stop anybody from using longer names. The real problem is that the documentation doesn't explain this, so a lot of people think they have to limit to 7 chars. By the way,since (2) implies (1), there is really only one rule. The new Internet domain naming scheme will allow much longer names, e.g. cbosgd could be d.osg.cb.btl.uucp if we wanted. (How this interfaces with UUCP is unclear - hopefully the answer is the same way the automobile interfaced with the stagecoach.) And if you're all worried about having to type these long names, that's what macros are for. Let your mail software do it for you.
ld (02/17/83)
Since Mark Horton choose to publically reproach my statement regarding site names...I choose to publically defend it. I promise that I will not continue this discussion on the net (hint, hint, Mark? Mail to me. I will publish a retraction if it is warranted). I would have just mailed to Mark myself, but I felt that this argument would be of some value to uucp/news administrators. To recap (because of the strange ordering netnews uses when handing postings to the reader) the original posting and Marks reply , I stated: When administering...uucp...three rules...in naming said site. 1) Site names must be unique. 2) Site names must be the same length. 3) Site names must be no more than 7 characters. Mark Horton replied: Close but no cigar. The real rules are 1) Site names must be unique. 2) Site names must be unique in their 7 characters. There is a bug in many UUCP's... As long as the bug exists in "many" (most?) uucp's, then the 3 rules I stated are correct. As long as there are sites that only have binary licenses (legally), then they should be made aware of this bug. That was the intent of my original posting. Once they are made aware of the bug, then it is a feature (and I get a cigar from those binary sites I inform). Posting the fix does no good to non-source sites. Remember, we do not all work in that great phone booth in the sky. Since I began my original posting with "When administering a MULTI-NODE-STAR installation, ..." and "... do not NECESSARILY mix ..." (emphasis added), I was addressing the article to a certain type of site/administrator and I was indicating that it would probably work anyway. The entire article was prompted by my having to search for the prefix routine (yes, I have a source license) and verifying what I suspected when mail was being misdirected. Without the source, I would have been at a loss to interpret the uucp LOGFILE. Hence, I can save some other uucp slave (there are no uucp masters) from the same fate. Until the definitive uucp source that has <NO> bugs (ha!) is collected and distributed to all sites...!? The question is, will Bell or Berkeley distribute the best version of uucp first? While it is true that site names can be longer than 7 characters, it is not good practice to violate this rule (just as it is not good practice to violate the rule in C). Since I do not want to be known for suggesting a violation of rules (I am known for a lot worse things), I deliberately posted rule 3. Understand that, while I am not the most experienced uucp hacker, I am not in inexperienced twit either. My error in posting the original statement was one of omission. I did not tell anyone why I felt the three rules existed (the bug). If I had, then Mark probably would not have replied. Finally (pant pant), I look forward to the Internet domain naming scheme, but Internet may still use uucp, and the problem will not go away. Thus, the mail software may still expand d.osg.cb.btl.uucp into ..!cbosgd!.. QED Note to Mark: Much has been posted about emotions coming across the net incorrectly. I just want you to know that this is a non-belligerent reply. Uucp hackers need all the friends they can get (we lose a lot of friends explaining why their mail was dropped on the floor). Larry Dwyer ucbvax!hpda!ld
silver (02/19/83)
Although you can't tell, I really work on one of the HP machines with a dash in the name. We're talking about removing it. So why is it there? Remember Silverstein's Principle: "Any Standard Is Better Than None." ... we just followed the precedent! Alan Silverstein csu-cs!hp-esd!ajs