root@hobbes.UUCP (02/20/87)
> Dave Stewart <2436@sequent.UUCP> responds to: | Dave Taylor: <1311@hplabsc.UUCP> | We could do perfectly well by eliminating the whole concept of backbone sites. | > The bigger reality in my mind is that this topology reduces article > propagation time. Consider a "truly anarchic" Usenet, full of nodes > with fanout of three or less and with long distance charges as low > as possible. The result would be that news would take much longer to > get about. | I think the net would be *better* if we didn't have backbone sites at all. | As far as using phone bill cost as a reason for anything, then my reply | is "then let's change the topology!". First, some history of another net which started off "truly anarchic". FidoNet, circa 1984: The topology of FidoNet is basicly that of an amorphus blob. Every site has a direct (dialup) connection to *every* other Fido in the world. The philosophy is to "provide micro-computer users with an *inexpensive* network using modems and voice-grade phone lines". The bottom line is that the phone costs will be born *by those sending messages*. Think about that. That means that if YOU want to send a message to every other system on the net, you would have to call every other system on the net. In 1986, that would be > 1200 phone calls, most of them long distance. Remember, no backbone, no news broadcasting... After all, why should *I* pay for your email? Circa 1985: Hubs were then introduced to this blob. They were transparent to the user in that your local machine talked to your local hub, which sent the message to your receiver's hub which sent it to your receiver. You then remunerated your hub for the phone charges. Since the hubs could bundle messages from all of its local nodes, the costs went down. Some hubs could use company WATS lines after hours, so their charges were *really* low. Soon, the question came up about routing mail between Fido's such that each link was a local call. It was determined that even if the routing software could be written to avoid loops and dead machines, the propagation delays would be enormous! 15 to 20 local calls per 100 miles... SLOW! Besides, wouldn't it be better to use 'that hub over there which has WATS lines...' There we are back at backbones (I mean hubs). Why? Because of the bottom line: it is *much* cheaper to use hubs which can afford or justify the phone connections than it is for you to make the calls yourself. Back to Usenet: Sorry if I'm being blunt. We are taking advantage of the generosity of those we call backbone sites. There is nothing which says that seismo or cbatt or decwrl or ihnp4... must provide you with free uucp mail and news forwarding! These sites have volunteered to provide this service to us. In exchange for the -free service- provided by these hosts, they have earned (bought?) the right to have a say over how we use them. The alternative is always there to not use the backbone hosts and to call direct or to setup your own independent topology. But, I'll bet that if you do, you will *end up* using some sort of backbone, just with other machines. |As far as using phone bill cost as a reason for anything, then my reply |is "then let's change the topology!". | I don't think it is a good justification for having more limited newsgroups, |more moderation, or whatever else is "in the works". Ok, what do you suggest? The people paying the bills have said that they *must* keep their phone costs down or risk even greater clampdown on THEIR usenet interaction. They have said that one way for them to do that is to control the creation of newsgroups. Moderated groups cost them (us!) less than unmoderated ones. At last count there were 310 newsgroups in my active file. My God! I'd think that there would be something already there which would satisfy most desires. If not, let's see what happens: MicroPort announces a full port of Unix System5 release 2 to the IBM AT. The ibm-pc groups hear about it. There is some interest, but not enough to warrent a new group, yet. So, a mailing list is formed. If the interest grows, I'm sure that a moderated group would be the next logical step - someone to keep tabs on things and to add coherance to the group. Later, if the moderator gets swamped, an open group would work. Let's see what's out there... Comp.bugs.sys5 and mod.os.unix. Gee, we could fit into these with no real problem! Andy Tannenbaum publishes a book on a 'Unix like OS for the PC'. The comp.sys.ibm-pc groups (moderated and unmoderated) start filling up with Minix articles. Poof! A new newsgroup comes up - comp.os.minix. No real problem, there was a need, it was filled. See, the system works; in fact, it is working better than it was just a year or so ago. The seams are starting to stretch, and the tailor is recomending prudence before the seat rips out and the pants fall apart. You can scream that the tailor is being dictatorial and arbitrary; it's your pants, you can do what you want :-) If you find a better pair, fine, everyone will want to know about it. If on the other hand, this pair falls apart before you have found a replacement, there may never be another. -john- Dave Stewart & Dave Taylor: If you have flames, as opposed to followups, please email. I don't mean to be uncharitable to either of you; I just have been active in a very anarchistic net, and I find that the strong backbone of Usenet is the main thing which makes this net something which is easy and painless to use. Would you post messages like you do if each one was billed to you for $100? (1000 machines X $0.10) Single Fido messages cost about $.25 each, $2.25 overseas. hobbes!plocher@unix.macc.wisc.edu <backbone> !uwvax!uwmacc!hobbes!plocher %s/spelling mistaks//g