Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ukma!husc6!ncar!woods From: woods@ncar.ucar.edu (Greg Woods) Newsgroups: comp.mail.misc Subject: Re: Short-circuiting bang paths Message-ID: <3144@ncar.ucar.edu> Date: 3 May 89 20:19:08 GMT References: <3048@tank.uchicago.edu> Reply-To: woods@handies.UCAR.EDU (Greg Woods) Organization: Scientific Computing Division/NCAR, Boulder CO Lines: 33 In article <3048@tank.uchicago.edu> matt@oddjob.uchicago.edu (Matt Crawford) writes: >I'm sure the quick and popular answer to this problem is "when the >listed first hop is a valid connection, take it." But I could counter- >argue that the mail exchanger for the target site was poorly chosen if >my own site is closer to the target than the exchanger is. This is another of the practical vs. ideal arguments. As maintainer of one of the sites in question, there is an MX record out there that points *.fidonet.org to my machine. The only service I perform for Fidonet is to convert domain addresses to bang paths. I do not have any physical connections to Fidonet at all. This means someone mails to user@whatever.fidonet.org, my mailer converts this to a bang path to the appropriate UUCP-to-Fido gateway (found from UUCP map data), and mails it out. From what I can see, there are over a dozen such gateways, and each gateway can itself gateway to dozens of Fidonet hosts. To do this the "right" way would require possibly hundreds of MX records. In addition to that, these MX records would have to be maintained by someone else who is providing name server service for fidonet.org (who may or may not themselves be connected to Fidonet). It isn't likely that they could find someone who would volunteer to provide this service under those conditions. Nor would I be willing to do routing if I had to maintain all the info myself. As it happens, they maintain their own info by publishing UUCP map entries for the various hosts. This means that those who benefit do the real work. It's practical this way and impractical to do it any other way. Then we get back to the religious war over whether aggressive rerouting is a good thing or not. No point in arguing it; there are lots of well-respected people on both sides of the issue. I got around the problem by simply deleting oddjob from my list of connections (by putting the -d oddjob flag on pathalias, which will prevent the generation of routes through oddjob from this site; the physical link still exists). End of mail loop. --Greg