mjc@cad.cs.cmu.edu (Monica Cellio) (11/04/85)
Ok, someone tell me the flaw in the following reasoning: If only 1 person wants to see a long source posting, it is obviously better for the net if the poster mails it to that person. If 10,000 people want to see it, it is obviously better to post it. I think there's a break-even point that can be determined; if fewer people than that magic number want to see it, it is better to mail, and if more than the magic number want to see it, it's better to post. Now obviously there are factors that can't be accounted for easily; for instance, if magic-number + 1 people want to see it and they're all on the same machine, it's still better to mail it (to one of them). Let's discount these factors for the moment (or is discounting these the flaw?). For the moment, let's say the poster has enough common sense to deal with these situations. What I propose is that for any "long" source posting, (what's long? I'm not sure. Certainly anything that's going to have to be broken into more than one piece.) the poster sends a note to the appropriate source group saying, "I have this available." After two weeks, he either posts or or starts mailing it out, depending on the number of replies he gets. Optionally, he can also post a list of people who have it back to the source group, so if someone sees the message late and someone closer to him than the poster has it, he can get it from that person. You will notice that I haven't proposed a method of determining the magic number. That's because I don't know how to. But I think it is possible for someone to sit down with the map and some cost estimates from the backbones and work out a rough guess. (Hey, backbones: How many mail messages to random destinations coming through your site equal one posting? What are the odds that a random mail message *will* come through your site?) Comments? -Dragon -- UUCP: ...ucbvax!dual!lll-crg!dragon ARPA: monica.cellio@cmu-cs-cad or dragon@lll-crg