jww@sdcsvax.UUCP (Joel West) (10/07/85)
In article <273@graffiti.UUCP>, peter@graffiti.UUCP (Peter da Silva) writes: > The only problem I see with the bang scheme is that sites > are ignoring it. If they gave it precedence over ARPA and other non-bang > syntax it'd work fine... > > If (the address contains BANGS && > I am adjacent to the next site in the path) > Send it to them; > > Use whatever heuristics you want to handle ambiguous paths, but if > you get a UUCP message that is addressed to the site next door to you, DON'T > look any further! I can't agree. This may be a religious debate, but the use of "Bang precedence" is what causes all sorts of ambiguities in the system. If I say pyramid!bugs@sri-unix.ARPA I want it delivered to the system that is one hop away from sri-unix on the ARPA net. This is a standard, recognizable address that all ARPA sites and many UUCP sites can handle. Now if I'm at gould9!cacimv!joel@NOSC.ARPA, so what? Just let me take the address pyramid!bugs@sri-unix.ARPA and pass it on to gould9 as is. Gould9 does the same towards NOSC, even though it is a UUCP-only site without ARPA access. [Watch my lips: THE ADDRESS IS NOT REWRITTEN] In fact, NOSC can take the address as is, and hand it to SRI-UNIX as is. [THE ADDRESS IS NOT REWRITTEN]. This is the magical part about domains: they present an isotropic address space. (Look that up in your Funk & Wagnalls!) No matter where you stick the mail message in, properly handled, it will arrive at its destination. But of course, if one site decides to turn this standard on its head, all hell breaks loose. Joel West CACI, Inc. - Federal (c/o Gould CSD) {ihnp4,cbosgd,decvax!sdcsvax}!gould9!joel gould9!joel@NOSC.ARPA (an unambigous address)
jer@peora.UUCP (J. Eric Roskos) (10/15/85)
> I can't agree. This may be a religious debate, but the use of "Bang > precedence" is what causes all sorts of ambiguities in the system. Actually, it's not a "religious debate"... it's a debate between those who advocate minimizing the impact on sites unconcerned (even unaware) of this whole issue, and those who think everyone should change at once for the common good of all. > This is the magical part about domains: they present an isotropic > address space. This is not a property of domains per se; that is another of the points where this debate becomes confused. This is a property of having a routing language in which you can specify domains without the domains being tied to a higher-precedence token in the routing language (which would limit where the domain names can be used in the strings). It's ironic that both sides here are arguing for a language in which all tokens have equal precedence in order to solve the above problem, but one group wants to give one token (@) higher precedence, then say "don't use it," whereas the other wants to give it equal precedence. That's the only point of disagreement. I think everyone agrees that extending the UUCP routing language to include domain names without their being tied to the "@" token is a good idea. Thus I don't think anybody here who is advocating !-precedence disagrees with the problem or solution you've stated. I certainly agree, and I suspect many others would as well, that writing an all-! string for this route is the solution to this problem. However, it is an *extension* to the existing syntax that need be implemented only at sites that wish to serve as "gateways" or "bridges" between networks, and it does not *require* you to have @-precedence. Having the syntax like this: ...!site1!site2.dom!site3!... is not at all incompatible with having strings like this: ...!site1!site2.dom!site3!joe@joesvax.ARPA in which ! still has precedence, as it does now, and which "accidentally" work at present. The only difference in the "other" approach is that it proposes that because a few sites currently, on their own, give "@" precedence already (those that route everything through Sendmail), then everyone should. But my argument is that those sites giving "@" precedence was an error -- it certainly resulted from installing software incompatible with the existing non-Berkeley UUCP sites -- and that the error should be corrected at those sites, not by making everyone everywhere do it too, especially when it doesn't give you any advantage. Now, this brings in the second problem you combined with the first one, how to specify the addresses in the user interface. And I agree there that you should give @-precedence, as long as you also allow the user to write addresses with all !'s in them for use with special cases. Because, as I said many times the time before, RFC822 is a standard for the format of the message headers, the things the user interacts with, whereas the routing language is a *separate* language altogether. The only case where it is not is where people run everything through Sendmail, regardless of whether or not they need to; and *that* is indeed a religious debate ("I like sendmail a lot and it gives me features I need and I don't have time to make it work right with UUCP so I'll just convince everyone to change the way UUCP works to suit sendmail"). My next posting will be on incompatibilities between MH.5 and non-Berkeley, post-Edition-7 Unix*-like systems. I.e., I am not going to continue restating the above again and again... please reread the above thoroughly before restating the same thing that has been stated repeatedly before, because I'm not going to restate MY position again, unless some *new* argument is presented to counter it. -- jer *Unix is a trademark of AT&T. -- Shyy-Anzr: J. Eric Roskos UUCP: Ofc: ..!{decvax,ucbvax,ihnp4}!vax135!petsd!peora!jer Home: ..!{decvax,ucbvax,ihnp4}!vax135!petsd!peora!jerpc!jer US Mail: MS 795; Perkin-Elmer SDC; 2486 Sand Lake Road, Orlando, FL 32809-7642
kre@munnari.OZ (Robert Elz) (10/17/85)
As best an I have been able to determine, no-one is advocating that all sites should be required to give @ precedence in rmail commands (uux message delivery). A site that doesn't know any @ mailers, just runs v7 (or Sys V) mail, and only knows uucp is, and is always going to be, quite free to just lop off the first "host!" and send mail to that host, regardless of what is in the rest of the string. But just as it is unreasonable, and impracticable to expect sites to give '@' precedence at the uucp transport level, its also impracticable to expect that every host in the world is ever going to suddenly stop giving '@' precedence either. It just won't happen. What all that boils down to, is that unless you know what every host on the path that you are using to route mail will do with an '@' at the uucp transport level, then its unsafe, unwise, and downright foolhardy to use one. Its also NEVER necessary. As long as there is ONE host that will properly handle ...!host.dom!... addresses, that's all that you need. It may be expensive, and unwelcome, but you could use uucp routing to get to that one host, and then have it do the right things. But we know that there is more than one such host already, so you actually get a choice. There is NO requirement that every host that could possibly be a gateway between the internet world and the uucp world do this translation - if the one that happens to be cheapest for you to mail to doesn't, then hard luck. (You could always politely request that they change things). This "issue" that has been flamed about off and on for ages now is really a non-issue. '@' at the uucp transport level is unsafe. Without requiring people to change their software, it will remain unsafe. Sending mail with unsafe addresses is not a nice thing to do, and is unnecessary. Can we end this trivia please? Robert Elz seismo!munnari!kre kre%munnari.oz@seismo.css.gov ps: a reason that some sites like to give '@' precedence, in any mail address, is that that means that an address at that site *always* means the same thing. Its not important how the address happened to reach the site. Given mail address X, this is how we will treat it. Personally, I think that is a laudable aim, but I agree, this *is* a religious issue. Arguing it is pointless.
peter@graffiti.UUCP (Peter da Silva) (10/21/85)
> In article <273@graffiti.UUCP>, peter@graffiti.UUCP (Peter da Silva) writes: > > The only problem I see with the bang scheme is that sites > > are ignoring it. If they gave it precedence over ARPA and other non-bang > > syntax it'd work fine... > > > > If (the address contains BANGS && > > I am adjacent to the next site in the path) > > Send it to them; > > > > Use whatever heuristics you want to handle ambiguous paths, but if > > you get a UUCP message that is addressed to the site next door to you, DON'T > > look any further! > > I can't agree. This may be a religious debate, but the use of "Bang > precedence" is what causes all sorts of ambiguities in the system. Why? > If I say > pyramid!bugs@sri-unix.ARPA > I want it delivered to the system that is one hop away from sri-unix > on the ARPA net. This is a standard, recognizable address that > all ARPA sites and many UUCP sites can handle. That's fine. If you're not adjacent to pyramid then, by the algorithm above, you use local mailing heuristics which presumably use ATSIGN precedence. Everything works out fine. I don't see your problem. > Now if I'm at > gould9!cacimv!joel@NOSC.ARPA, > so what? Just let me take the address > pyramid!bugs@sri-unix.ARPA > and pass it on to gould9 as is. Gould9 does the same towards NOSC, > even though it is a UUCP-only site without ARPA access. > > [Watch my lips: THE ADDRESS IS NOT REWRITTEN] Did I say anything about rewriting addresses? Since gould9 doesn't have any direct UUCP links to pyramid it should go by local rules & pass it off to whoever can handle it. Sending it off to .ARPA is a reasonable action. Watch MY lips: IF YOU ARE ADJACENT TO THE NEXT UUCP SITE IN THE PATH, THEN ACT LIKE A UUCP SITE. OTHERWISE (as in this case) DO WHATEVER YOU WANT TO. > In fact, NOSC can take the address as is, and hand it to SRI-UNIX > as is. [THE ADDRESS IS NOT REWRITTEN]. > > This is the magical part about domains: they present an isotropic > address space. (Look that up in your Funk & Wagnalls!) No matter > where you stick the mail message in, properly handled, it will > arrive at its destination. But .ARPA isn't a domain. > But of course, if one site decides to turn this standard on its > head, all hell breaks loose. In UUCP the standard is BANG routing. As you say, if one site turns it on its head, all hell breaks loose. I still don't understand what you don't like about my article, apart from the fact that I like bang precedence in certain carefully defined areas. What's the problem? We seem to be in basic agreement.
peter@graffiti.UUCP (Peter da Silva) (10/21/85)
What about sites that insert @ and even % in return paths, so that there is no way to figure out how to reply to a peice of mail? What am I supposed to do with a path with an @ and 2 sites I know to have ARPA access in it? Send 2 copies of my reply & hope for the best? Suggestion: when you put something back into .UUCP from .ARPA, make sure it's in BANG syntax.