Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!usc!ucsd!helios.ee.lbl.gov!ncis.tis.llnl.gov!lll-winken!uunet!mcsun!ukc!dcl-cs!aber-cs!thor!pcg From: pcg@cs.aber.ac.uk (Piercarlo Grandi) Newsgroups: comp.mail.uucp Subject: Re: Imminent death of UUCP Zone predicted Message-ID: Date: 22 Jul 90 19:05:54 GMT References: <1990Jun28.164938.23367@DSI.COM> <3008.268b1e9a@mccall.com> <26669@ditka.UUCP> <7871@lynx.UUCP> <100@raysnec.UUCP> <269B82AE.415E@intercon.com> Sender: pcg@aber-cs.UUCP Organization: Coleg Prifysgol Cymru Lines: 54 X-Local-Date: 22 Jul 90 12:05:54 PDT In-reply-to: amanda@mermaid.intercon.com's message of 11 Jul 90 19:49:02 GMT In article <269B82AE.415E@intercon.com> amanda@mermaid.intercon.com (Amanda Walker) writes: In article <100@raysnec.UUCP>, shwake@raysnec.UUCP (Ray Shwake) writes: > As to the address formats, there are really only two that could > remotely be considered "intuitive", viz: > > user@sitename (specified user AT specified sitename) > site!...!site!user (reach user by following this path) Fine so far. However, the second is not an address. It is a route. There are many mailers that blur the distinction, but it is there nonetheless. This is not entirely right. This is a *relative* name (interview to Peter Honeyman, some issue of Unix Review of a few years back). Used as a name, it is like a japanese address: a!b!c!u says that I want to send the mail to user 'u' at the host called 'c' which is a neighbour of 'b' which is a neighbour of 'a'. It *does not* imply that the mail will pass thru 'a' and 'b' on its way to 'c', even if this is the most common intepretation. Actually, if you do not do aggressive rerouting, this may well be used as an underspecified route, e.g. something where the current host has all liberties in choosing which path to use to get to 'a', and then 'a' chooses by which path reach 'b', and 'b' gets to 'c' by any path. This is my favourite compromise between relative naming and source routing, and does blur the two a bit, but conveniently. Note that strictly speaking the relative address, even if used as a route, ought not be trimmed as you get nearer the destination, because it is a *name*, and moreover it is kind of 'absolute'; it identifies host 'c' near 'b' near 'a', and it is only ambiguous if there is another 'c' near a 'b' near an 'a'. In a sense this gives unique host naming by pattern matching instead of rooting. Let me repeat my impossible dream of the Internet switching to bang addresses, possibly underspecified source routing, and relative naming... -- Piercarlo "Peter" Grandi | ARPA: pcg%cs.aber.ac.uk@nsfnet-relay.ac.uk Dept of CS, UCW Aberystwyth | UUCP: ...!mcsun!ukc!aber-cs!pcg Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk