Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!rpi!uupsi!sunic!nuug!ifi!enag From: enag@ifi.uio.no (Erik Naggum) Newsgroups: comp.protocols.tcp-ip.domains Subject: Re: CS top-level domain and its impact on the UK? Message-ID: Date: 29 Jul 90 20:44:55 GMT References: <1990Jul23.065901.8741@ifi.uio.no> <1990Jul28.235750.20841@kth.se> <1990Jul29.103214.14864@robobar.co.uk> Sender: enag@ifi.uio.no (Erik Naggum) Organization: Naggum Software, Oslo, Norway Lines: 153 In-Reply-To: ronald@robobar.co.uk's message of 29 Jul 90 10:32:14 GMT Originator: enag@slembe.ifi.uio.no Nntp-Posting-Host: slembe.ifi.uio.no Lines: The saga continues... In article <1990Jul29.235750.20841@kth.se>, Per Andersson suggests: >> Let them be known as foo@bar.baz.janet or something. In article <1990Jul29.103214.14864@robobar.co.uk>, Ronald S H Khoo replies: > I hope I'm right in reading sarcasm into this statement. I thought we got > away from network based addresses LONG time ago. Well, not yet. Remember BITNET. UUCP is still out there. It could be a good solution to introduce JANET as a fake-domain until you guys clean up your act and get it right. > Besides, you're mixing up two separate issues there: Excuse me, but I think you're the one who's mixing up things around here. Your particular problem is that you have nonfunctional gateway software, and you pretend to make this a problem at the other side of the gateway. Since I'm at the other side, and care about doing things right at this side, I get pissed each time some UK idea comes up to "adapt" the world to their silly ways. > The technical issue: the coloured book transport mechanisms officially > used on JANET uses reversed names. That's TRANSPORT. UUCP transport is > far worse, and yet it is possible to Get Things Nearly Right with UUCP. Wrong. Domain names are properly referred to as _session_ mechanisms, if you want to use that contrived ISO model. I don't know the details about your colored books, and I don't want to, but I suspect that at the transport layer, you talk about some X.121 address or at least something to stuff into the address block of the call request packet of your very advanced X.25 network. Likewise, we don't use domain names at the TCP-level, we translate the domain name into a 32-bit IP-address before TCP gets involved. Domain names inhabit the app- lication layers in the Internet protocol suite. There is nothing wrong about using domain names at the application layer that translate to UUCP node names somewhere at the transport layer, further enhanced by fancy telephone numbers. People do that all the time. There is even a draft out which considers ISDN addresses for Internet transport services (TP/IX). The problem with this insipid reversed name style is that it is up to the applications to do it right, and it wouldn't have been a problem if you UK types could get sufficiently intelligent designers from abroad to write your software. Of course, since you're backwards, and your users won't exactly understand that all by themselves, they will use Internet domain names for addressing purposes. So, you can choose between a major operation to educate the users to understand that UK is one way, the rest of the world is the other way, or a major oper- ation to get this silly JANET convention killed and safely buried. Nobody will mourn its passing. > The current situation is in fact only minimally sub-optimal. NRS > addresses ARE valid Internet addresses -- The Internet domains under > .UK belong to the NRS, that is all. And yes, Piercarlo, they _are_ > Internet domains. Ah, I'm sorry to reveal a maximally sub-optimal and important distinction between NRS names and Internet domain names (not _addresses_, damn it!). You've probably heard it before, so it shouldn't be new to you: NRS names are _reversed_ relative to the order used by the DNS. Novel, eh? > OK, so 99.9 % of the domains under .UK are just MX'd behind either the > nsf.ac.uk or mcsun.eu.net (like me) but that's neither here nor there. > Lots of North American domains are similarly MX'd behind uunet.uu.net, > et. al. What's the difference ? The difference is: They're REVERSED! God, this is getting to be really funny. > Simple. The NRS scheme requires that ALL addressable hosts are > registered in the database. The correct solution is to try the reversed > addressed address for an EXACT match in the NRS database, and if it > isn't there, then it isn't a valid NRS address, SO USE IT THE RIGHT WAY > ROUND. Some JANET-type oughta write up an RFC for this sorta thing :-) I hope no JANET-type will ever attempt such a horrible thing. This is a JANET-local problem, and should properly concern not a single person off your particular island. There is no guarantee that FOOBAR.CS (Internet) and FOOBAR.CS (JANET) will not exist at the same time. The correct way to do it is not to ask the Internet first, either, but to drop the JANET convention, or to successfully hide it from the rest of the world. The latter is a truly major undertaking (as witness the current discussion), while the former is routine in comparison. > The heuristics are similar to those you need for matching incomplete > local addresses. After all, a hypothetical jerry@complex.math.foo.edu > who wanted to mail freddy@gizmo.cs.foo.edu might quite reasonably say > that "mail freddy@gizmo.cs" should work as he expects, and only if > gizmo.cs.foo.edu doesn't exist does he expect the mailer to look in > Czechoslovakia. The Internet decided to deprecate this practice about two years ago. It is no longer relevant, precisely because of the problems that arise from its usage. The Host Requirements RFC mandates use of Fully Qual- ified Domain Names in all instances. This does not require heuristics or randomly-lose options, only simple algorithms. > Whether it does or not should be a LOCAL policy issue and not impinge on > the Internet at large. I would personally prefer not to have lost the > ability to mail the REAL gizmo.CS and REQUIRE FQDNs even internally > ("so set up a mail alias" would be what I'd say) but that's a *LOCAL* > policy issue. Wrong. It ceases to be a local policy issue the day some unsuspecting individual discloses his address to another user. This is also why that silly JANET convention has become a global problem: Some unsus- pecting users will give out their JANET-version address to some other unsuspecting users, and interoperability suffers, at the cost of frustrated users and postmasters wasting their time at both ends. The rules for when it would be appropriate to use a local, abbreviated domain name are very complex. It's not a string comparison, to put it that way. > The kludge of trying to accept addresses either way round globally > is just a bad kludge. But it did mean that no decisions needed to be > made, I guess. Now the kludge is starting to get strained, perhaps > someone will have to do something. I must admit I'm glad it's not me. Someone, that is, from United Kingdom. The problem is that all too many of your countrymen are so fucking glad it's not them that it's never going to happen. And the really sad thing is that this stupidity is wasting time all over the place. Users can't get mail to work properly, signatures, letter heads and business cards advertise inappropriate addresses, postmasters get bugged by their users, people think that electronic mail in general doesn't work, manual authors devote a whole chapter to perverted addressing forms, lots of network people waste their time trying to cope with the problem, and hundreds of articles are written to denounce the silly Brits, and some silly Brits making a successful attempt to prove that there is no hope to get them to understand and change their ways. The person responsible for this idiotic decision should be flogged. Sometimes I think we would do lots of people a favor by _sinking_ the entire island. I mean, taking out a five mile thick horizontal slab of land below sea-level and move it somewhere it could earn a more useful keep than it does today. This would give real substance to the Atlantis myth, excepting of course that Atlantis was supposed to be a nice place with streets of gold. That will have to remain a myth. I have a map of Europe on my wall that has a nice blue color off the northwest coast of France. Beautiful. -- [Erik Naggum] Gaustadalleen 21 +47-256-7822 N-0371 OSLO; NORWAY +47-260-4427 (fax)