vixie@decwrl.dec.com (Paul Vixie) (11/18/88)
# $BEGIN PIPEDREAM$ # Of course, the real solution would be to see everyone on the Internet. # Maybe in our lifetimes? # $END PIPEDREAM$ See Subject: line. To _me_, "on the internet" means, in the context of mail, that you have either: [1] an A record in the DNS and an SMTP server, or [2] an MX record in the DNS which points to a host with an A record in the DNS that has an SMTP server. I am open to alternative definitions. Because by this definition, everybody could be on the internet and 90% of all links could be over UUCP. Remember, $35 to UUNET, paid one time only, gets you a domain name and an MX record. -- Paul Vixie Work: vixie@decwrl.dec.com decwrl!vixie +1 415 853 6600 Play: paul@vixie.sf.ca.us vixie!paul +1 415 864 7013
kurt@pprg.unm.edu (Kurt Zeilenga) (11/18/88)
Paul Vixie writes: >I am open to alternative definitions. Because by this definition, everybody >could be on the internet and 90% of all links could be over UUCP. Remember, >$35 to UUNET, paid one time only, gets you a domain name and an MX record. To me, on the Internet (note the capital 'I') implies you can hold some form of IP conversation with SRI-NIC.ARPA (or some other major ARPAnet/MILnet/NSFnet host). Being on an internet implies you talk IP (even if just to yourself :-). The terms I prefer to use in the context of mail is "directly attached" "indirectly attached" to the Internet. What we really need to do is get MX records for those hosts/sites not directly attached and blow off the UUCP maps all together.... -- Kurt
vixie@decwrl.dec.com (Paul Vixie) (11/18/88)
# To me, on the Internet (note the capital 'I') implies you can hold some # form of IP conversation with SRI-NIC.ARPA (or some other major # ARPAnet/MILnet/NSFnet host). Being on an internet implies you talk IP (even # if just to yourself :-). Well, my Symmetric 375 can carry on an IP connection with itself on lo0. I don't see this as a useful definition. I see the :-), but remember that I said "in the context of mail", and in that context, according to my usage of "on the internet", having an MX gets you on the internet. In fact, _in a mail context_, the ability to do IP with SRI-NIC.ARPA is not meaningful: you have to have an SMTP server running, which is not contained in that def'n. # What we really need to do is get MX records for those hosts/sites not # directly attached and blow off the UUCP maps all together.... That won't work. Every UUCP-connected site can have an MX in the DNS, and those sites will _still_ want UUCP maps to talk to _eachother_. Also, not all MX->client connections are direct. We don't want all traffic to have to have at least one IP hop in it, do we? -- Paul Vixie Work: vixie@decwrl.dec.com decwrl!vixie +1 415 853 6600 Play: paul@vixie.sf.ca.us vixie!paul +1 415 864 7013
rsalz@bbn.com (Rich Salz) (11/19/88)
In <23667@pprg.unm.edu>, Kurt Zilenga <kurt@pprg.unm.edu> writes >To me, on the Internet (note the capital 'I') implies you can hold some >form of IP conversation with SRI-NIC.ARPA (or some other major >ARPAnet/MILnet/NSFnet host). Okay, this sounds fine. > Being on an internet implies you talk IP (even >if just to yourself :-). Unh, this I disagree with. Two Sun's connected by SLIP ain't an internet. It's a LAN. If you've got Ethernet segments, or a gateway box, or something similar, then you've got an internet, if not you've got a LAN. An internet is an "interconnected network." Any large campus will have an internet, and any small shop will have a LAN. Note that most folks, as soon as they connect to the Internet, will have an internet, because they'll connect at one point and have that machine gateway into their LAN. >The terms I prefer to use in the context of mail is "directly attached" >"indirectly attached" to the Internet. What we really need to do >is get MX records for those hosts/sites not directly attached and blow >off the UUCP maps all together.... Well, kinda. It's hard to do. The handful of Internet folks willing to be MX forwarders must be willing to take on hundreds of connections (how many does UNM forward for? :-). Okay, this is an exaggeration. The other downside is that an it tends to mean that there's ONLY ONE mail path to a site. With UUCP, since I can specify routes, I can route around problems and get my mail through. With domains, I can't. The other alternative is "domain parks." An Internet site (eddie.mit.edu) acts as the MX forwarder for a domain (zone1.com) and it turns over all zone1.com mail to a largish UUCP site (mirror). This system works, as the real-life example in parens shows, but it's been less than wildly successful. I expect this is because of political problems people have with the UUCP project. /rich $alz -- Please send comp.sources.unix-related mail to rsalz@uunet.uu.net.
avolio@decuac.dec.com (Frederick M. Avolio) (11/19/88)
In article <56@gnome6.pa.dec.com> vixie@decwrl.dec.com (Paul Vixie) writes: > .... We don't want all traffic to have to >have at least one IP hop in it, do we? Well, actually many people say yes... Think about it a bit. Every UUCP site with an MX record. Every one within a hop or two of an Internet host. Would cut down on the UUCP maps. Fred
kurt@pprg.unm.edu (Kurt Zeilenga) (11/19/88)
Me: >The terms I prefer to use in the context of mail is "directly attached" >"indirectly attached" to the Internet. What we really need to do >is get MX records for those hosts/sites not directly attached and blow >off the UUCP maps all together.... All I was trying to say was that I wish UUCP and the UUCP maps would just fade away. That wish won't come true. The folks which are stuck using UUCP are just that, stuck with it. Luckily, the price of connecting to Internet is dropping quickly. But don't get me wrong, UUCP has and still does serve a purpose, I just would like to move to newer and better things.... rsalz@bbn.com (Rich Salz) writes on my statement above: >Well, kinda. It's hard to do. The handful of Internet folks willing to >be MX forwarders must be willing to take on hundreds of connections (how >many does UNM forward for? :-). Okay, this is an exaggeration. You're right, it is an exaggeration. > The other downside is >that an it tends to mean that there's ONLY ONE mail path to a >site. What? The Domain Name System (DNS) supports multiple MX records for any host or domain wildcard. They do have metrics. DNS mailers DYNAMICALLY select the "best" one available, avoiding down or unreachable exchangers. Even if your statement were true, I would rather heavily rely on one link from a site to its primary exchanger then rely on the many links in a UUCP route. This is because a site has direct control over the state its link to its primary exchanger. Everybody knows how map routing works, it just doesn't a good amount of time. The maps return poor or incorrect routes. Sometimes hosts along the path reroute things and then how good is your UUCP source route? Don't say it. I know, you can route around those sites, too. The key word here is YOU. Well, YOU can do it, I don't want to have to. > With UUCP, since I can specify routes, I can route around problems >and get my mail through. With domains, I can't. What? I do believe most DNS mailer support source routing... but I rarely have to use them. > /rich $alz One really nice thing about MX'ing is that the DNS doesn't care what the transport method is. The UUCP mapping project does care, UUCP sites only. Kurt
vixie@decwrl.dec.com (Paul Vixie) (11/19/88)
(Vixie, West Coast DEC) # [...] We don't want all traffic to have to # have at least one IP hop in it, do we? (Avolio, East Coast DEC) # Well, actually many people say yes... Think about it a bit. Every # UUCP site with an MX record. Every one within a hop or two of an # Internet host. Would cut down on the UUCP maps. For several reasons, I expect that UUCP or otherwise non-IP links will still exist for some time to come if not forever, since they are loose enough that even business competitors find it useful and secure enough to connect to eachother directly. _Directly_. Even in the case where an IP path exists between two machines, sometimes it may be quicker or otherwise desirable to move mail via UUCP. Given this, I expect the maps to continue to be useful for a long while yet, and I expect that some fraction of mail will always be deliverable along an ideal path that contains no IP hops. -- Paul Vixie Work: vixie@decwrl.dec.com decwrl!vixie +1 415 853 6600 Play: paul@vixie.sf.ca.us vixie!paul +1 415 864 7013