jerry@oliveb.UUCP (Jerry Aguirre) (10/22/85)
I was examining news articles for length of path and came across an interesting one. It follows the typical cross country path to get next door. Amazingly all this cross country hopping only took 3 days. Not bad except for the extra phone charges. I won't even try to figure out that %.@ junk on the end of the path. I will also leave it to someone else to figure out how the sender is brl-tgr when it didn't get there until 7 sites later. The interesting thing is the uucp path. It starts out in OR, hops across the country thru UT, MD, VA, MA, NJ, IL, NY, and then cross country back to WA, back to OR, and then more reasonably down to CA. One could find several points of optimization in this path and, no doubt, recommend several new connections. The most obvious though is athena->tektronix. This article took 21 hops, across country and back, to reach a site in the same state, the same city, and the same company. Hell, they are even in the same division of Tektronix. Please, someone tell me that the path has been munged and that it really didn't take that many hops to get here! Extract of article follows: > Relay-Version: version B 2.10.2 9/5/84; site oliveb.UUCP > Posting-Version: version B 2.10.2 9/18/84; site brl-tgr.ARPA > Path: oliveb!hplabs!tektronix!uw-beaver!cornell!vax135!houxm!ihnp4!mhuxn!mhuxr!ulysses!allegra!mit-eddie!genrad!panda!talcott!harvard!seismo!brl-tgr!tgr!utah-cs!admin!iti!athena!svc%utah-gr.UTAH-CS.ARPA.ARPA@BRL.ARPA > From: PA@BRL.ARPA> > Newsgroups: net.unix-wizards > Subject: Serial line problems on Sun > Message-ID: <1904@brl-tgr.ARPA> > Date: 4 Oct 85 06:30:32 GMT > Date-Received: 7 Oct 85 01:46:39 GMT > Sender: news@brl-tgr.ARPA Path site locations taken from mod.map data: svc%utah-gr.UTAH-CS.ARPA.ARPA@BRL.ARPA ?? athena Beaverton, OR iti ?? admin ?? utah-cs Salt Lake City, UT tgr ?? brl-tgr MD seismo Arlington, VA harvard Cambridge, MA talcott Cambridge, MA panda Concord, MA genrad Bolton, MA mit-eddie Cambridge, MA allegra Murray Hill, NJ ulysses Murray Hill, NJ mhuxr Murray Hill, NJ mhuxn Murray Hill, NJ ihnp4 Naperville, IL houxm Holmdel, NJ vax135 Holmdel, NJ cornell Ithaca, NY uw-beaver Seattle, WA tektronix Beaverton, OR hplabs Palo Alto, CA oliveb Cupertino, CA Jerry Aguirre @ Olivetti ATC {hplabs|fortune|idi|ihnp4|tolerant|allegra|tymix|olhqma}!oliveb!jerry
davest@lumiere.UUCP (Dave Stewart) (10/23/85)
In article <626@oliveb.UUCP> jerry@oliveb.UUCP (Jerry Aguirre) writes: >I was examining news articles for length of path and came across an >interesting one. > ... >The most obvious though is athena->tektronix. This article took 21 >hops, across country and back, to reach a site in the same state, the >same city, and the same company. Hell, they are even in the same >division of Tektronix. > ... >> Path: oliveb!hplabs!tektronix!uw-beaver!cornell!vax135!houxm!ihnp4!mhuxn!mhuxr!ulysses!allegra!mit-eddie!genrad!panda!talcott!harvard!seismo!brl-tgr!tgr!utah-cs!admin!iti!athena!svc%utah-gr.UTAH-CS.ARPA.ARPA@BRL.ARPA The athena here at Tek has no modems and is only connected to Usenet via ethernet to other Tek machines. I didn't think there were two athena's, but it looks like somebody took the name without consulting with the UUCP name registry - grr! Anyway, this should reduce the article's distance somewhat. Former news admin for Tektronix - -- David C. Stewart uucp: tektronix!davest Small Systems Support Group csnet: davest@TEKTRONIX Tektronix, Inc. phone: (503) 627-5418
rick@seismo.CSS.GOV (Rick Adams) (10/23/85)
The reason that path is weird is that the athena is NOT the athena at tektroinix, but an invalid duplicate in utah. The message was mailed to the arpanet unix-wizards mailing list and then made it to usenet through brl-tgr!seismo ---rick
jss@sjuvax.UUCP (J. Shapiro) (10/30/85)
> I was examining news articles for length of path and came across an > interesting one. It follows the typical cross country path to get next > door. Amazingly all this cross country hopping only took 3 days. Not > bad except for the extra phone charges. > > The interesting thing is the uucp path. It starts out in OR, hops > across the country thru UT, MD, VA, MA, NJ, IL, NY, and then cross > country back to WA, back to OR, and then more reasonably down to CA. Personally, I suspect that a lot of the cost of netnews occurs in just this fashion. Network links are cheap, but telephone lines aren't. I suspect that if we could get together a database of the kinds of links people have and the BellTel Cost/Kilobyte, along with phone numbers called, I suspect that we would find that there are a lot of duplicate routings (which is not necessarily bad), but more important, we would find that we could trivially encourage local sites to transmit to each other. The case above is a good case in point. Since most of the billing seems to be by phone, it seems logical to encourage that news travel over the most local connections. One of the original reasons for IHAVE/SENDME was that you could set up a primary feed, compile an IHAVE list, ship it to a secondary feed site, and get them to ship you whatever you didn't already have. While the three call cost is high if there is only a little news, it is well worth it if we can save 2 hours at the cost of two short calls. In short, I think it would make a lot of sense to reorganize the net propagation in accordance with phone area code and exchange. Does it seem likely that people would be willing to do this, and if so, is it worth my effort? I suspect that we would find that in an awful lot of cases, the long distance stuff could be made to run over already existing internal networks which are not nearly saturated. This would take a load off of some of the backbone sites in terms of phone cost, and would probably speed up overall propagation, because many more sites would be willing to call on demand. It would also make a lot of billing people happy. Last and best of all, it would make Ma Bell very Unhappy -); I am willing to try to write a script to compile this information from the existing net map files, but it seems to me that in many cases people have not given machine phone numbers, and therefore correlation by phone area code and exchange will be very difficult Can anybody think up a reasonable way of encourageing people to agree to trying this? Mail connections should be left alone, as companies have individual reasons for doing this. Jon Shapiro Haverford College -- Jonathan S. Shapiro Haverford College "It doesn't compile pseudo code... What do you expect for fifty dollars?" - M. Tiemann
friesen@psivax.UUCP (Stanley Friesen) (11/05/85)
In article <2489@sjuvax.UUCP> jss@sjuvax.UUCP (J. Shapiro) writes: > >I am willing to try to write a script to compile this information from >the existing net map files, but it seems to me that in many cases >people have not given machine phone numbers, and therefore correlation >by phone area code and exchange will be very difficult > >Can anybody think up a reasonable way of encourageing people to agree >to trying this? > well, first why not go ahead and put together that script and simply fudge the cases where there is no phone number, perhaps group them by state only? Then post at least a summary of the more important results and suggestions this produces as well as posting the script to mod.sources. I was going to do something like this by hand in my spare time, but a script would be better. -- Sarima (Stanley Friesen) UUCP: {ttidca|ihnp4|sdcrdcf|quad1|nrcvax|bellcore|logico}!psivax!friesen ARPA: ttidca!psivax!friesen@rand-unix.arpa
sewilco@mecc.UUCP (Scot E. Wilcoxon) (11/07/85)
In article <2489@sjuvax.UUCP> jss@sjuvax.UUCP (J. Shapiro) writes: >>... >> The interesting thing is the uucp path. It starts out in OR, hops >> across the country thru UT, MD, VA, MA, NJ, IL, NY, and then cross >> country back to WA, back to OR, and then more reasonably down to CA. > >Personally, I suspect that a lot of the cost of netnews occurs in just >this fashion. Network links are cheap, but telephone lines aren't. I Don't confuse overall cost of an article with the cost to each site of passing the article on to the next site. Each site pays only for its calls to its immediate neighbors. >... >... , we would >find that we could trivially encourage local sites to transmit to each >other. The case above is a good case in point. Don't need a giant database. Talk to the news coordinators at your neighboring sites and you can coordinate your regional/city net. We're doing it here in MN. If you don't have a statewide or citywide distribution yet, set one up with your neighbors. Remember also that the net is a social creation, and its humans are as important as its computers. >(explanation of IHAVE/SENDME) >... >of cases, the long distance stuff could be made to run over already >existing internal networks which are not nearly saturated. This would >... Some of those long distance hops might well have been over internal networks. We can't even know how many computers are sharing one external name, much less which are dialing through what phone/intercom circuits. Further assumptions are left as an exercise to the reader. Further complications will become apparent in later issues of "Dear Networld:" -- Scot E. Wilcoxon Minn. Ed. Comp. Corp. circadia!mecc!sewilco 45 03 N / 93 15 W (612)481-3507 {ihnp4,mgnetp}!dicomed!mecc!sewilco