gross@dcn7.ARPA (07/11/84)
I have a question or two for the mail wizards out there. Some of my concerns are probably old hat and I suspect you old-timers may be tired of answering them for us relative newcomers to net-hopping. Some of these questions, however, may actually be interesting. If there is a document (or 2 or 3, like the 'emily post' series), please feel free to point me that direction. Otherwise, I volunteer to collect and summarize all responses to produce such a document. All my questions evolved from the following situation: I occassionally send mail from our ARPANET host to PSUVAX1, which is a USENET and BITNET host but not an ARPANET host. I've used BERKELEY as my point of departure from Arpaland to Usenet, with an address like <U1!U2!...!psuvax1!user@berkeley>, which uses the familiar USENET source route as the 'local part' of the RFC822 'addr-spec'. Recently, PSUVAX1 made some changes and I began using an address like <user%psuvax1.bitnet@berkeley>, which uses the '%' convention and BITNET neither of which I am familiar with. Now I would like to send mail to BURDVAX, which I know that I can reach through PSUVAX1 via USENET. The smart money addressing choice was <U1!U2!...!psuvax1!burdvax!user2@berkeley>, however, I couldn't resist trying <burdvax!user2%psuvax1.bitnet@berkeley>, or <USEnetaddr % BITnetaddr @ ARPAnet>. I was pretty sure the second would fail because the '%' is nowhere to be found in RFC822, which means there's no formal operator precedence between '!', '%', and '@', which means all bets are off. Sure enough, BERKELEY took the first choice but flung the second back into my face. The second addressing choice (or some legal variant) is my favorite because if BITNET is the path of choice to get to PSUVAX1 from here (perhaps because it's not restricted to source routing like USENET), then I shouldn't be forced to use USENET for the whole trip simply due to addressing hassles. I also like it because it uses 3 dissimilar, non-traditional-arpa-internet nets (Anyone know a better route involving CSNET?). My smorgasbord of resulting questions are below. Again, I will summarize the results and produce a document that we can point guys like me toward in the future (unless, of course, one already exists). To avoid cluttering mailboxes with possibly old issues, replies should come directly to me (gross@dcn7), unless of general interest to the group. 1) I'm familiar with USENET (although a clarification between UUCP and USENET would be useful), but not with BITNET. Are there articles descibing it (like the 10/83 CACM or Sigcomm '83 Proceeeding articles on CSNET) or can someone give me a paragraph or two. 2) Since '%' (percent) it isn't part of the RFC822 grammer, what does it mean and what is the convention for parsing it? 3) Is the "usenetroute@arpahost" convention a Berkeley kludge or is it a recognized de facto standard of some sort? 4) Is there a permissable way to do the ARPA-to-BIT-to-USEnet addressing implied in the second choice above? In other words, is there an accepted way to parse a mixture of '!', '%' and '@' together. If so, how widely would it be accepted? 5) Who develops these conventions for non-Arpa-Internet inter- networking and how are they documented? 6) How do people determine efficient UUCP/USENET source routes? 7) Are Arpanauts guilty of a narrow worldview or should non-arpanoids accept the one true way? (I've asbestos lined my mailbox for this one, just in case.) This is another way of asking whether there may ever be an ecumenical 'son of 822' to codify some of this stuff? Awaiting Enlightenment, Phill Gross -------
lauren@vortex.UUCP (Lauren Weinstein) (07/13/84)
All bets are pretty much off when you try to gateway through more than one net, since the precedence rules are pretty much ad hoc and get confusing quickly beyond one gateway. However, experience has shown that in many cases the UUCP route [it's UUCP routing, not Usenet (the latter refers to netnews distribution paths] can be the most reliable for some purposes. For example, one mailing list that I'm currently on, via a site that supports both UUCP and CSNET, recently had to turn off all of the paths leading to and from the list via CSNET and instead use UUCP routing exclusively. Why? The CSNET software kept failing with the mailing list, causing all sorts of problems, while the UUCP code seems to be churning along just fine. They're still trying to figure out the problem. The moral: If you know a path that seems to work, don't spend too much time trying to find other ones. --Lauren--
jdi@psuvax1.UUCP (John D. Irwin) (07/19/84)
Speaking as psuvax1, first I'd like to clear up a couple of mis-
conceptions. First of all, our switch to 4.2 OBVIATED the need for a '%',
not required it. Currently if you can get mail to us through whatever
net that is 822 we will send it on (except for a teeny uucp problem we're
working on)
Thus djb@burdvax.UUCP.BITNET
sent to us would work. However, you have a problem in specifying HOW to
get it from Berkeley to us. Mark, any answers?
--
Spoken: John D. Irwin
AT&T: 814-237-5068
Nets: jdi@psuvax1.{BITNET,CSNET}
Uucp: {akgua, allegra, cornell, pitt, purdue, ihnp4, burdvax}!psuvax1!jdimark@cbosgd.UUCP (Mark Horton) (07/27/84)
There must be lots of machines that talk to both ucbvax and psuvax1 - cbosgd for example. But I confess I can't understand why djb@burdvax.UUCP.BITNET would be a useful address. If your system understands BITNET but not UUCP, then djb%burdvax.UUCP@psuvax1.BITNET or the more intended @psuvax1.BITNET:djb@burdvax.UUCP ought to work. From the ARPANET I think Berkeley will gateway to BITNET, from CSNET there's probably a gateway somewhere too. From UUCP it depends on what software you have and where you want to send the mail.
knutsen@SRI-UNIX.ARPA (Andrew Knutsen) (08/02/84)
The basic problem is that ARPA (the US gov't) hasnt declared any "official" domains other than ARPA. An official domain implys the existence of a well-maintained nameserver service, and this does not exist for uucp, or bitnet for all I know. Thus on the arpanet, ARPA must be top level, and subdomains must be in the "local part". Kind of a silly situation, but the good side is that it encourges the formation of the nameservers. Apparently things are coming along on the uucp and csnet fronts. AK