| Doc. #300 | February 1991 |
|---|
Network mail exchange depends not only on hardware arrangements of the systems involved but on the software packages, such as "mailers," that are responsible for routing incoming and outgoing messages on any one machine. Hardware as well as software environments will vary from system to system on the path between the sender and recipient of a message, and heterogenous e-mail address formats abound as a result. This User Guide is designed to help the typical UIUCnet user understand the e-mail address environment.
In Sections 4-8, specific address conventions between the three primary networks to which UIUCnet has e-mail access will be discussed.
The DNS is a distributed, hierarchical structure of names used by most Internet applications including e-mail, telnet and file transfer protocol (ftp). This naming structure reflects administrative levels of organization and/or geography on the Internet. The e-mail address jsmith@uxa.cso.uiuc.edu exemplifies a DNS-style address, and a simple analysis of this address will help explain the system and how it relates to mail.
A good place to begin in this case, as with many e-mail addresses, is at the "@" sign, since this delimiter typically occurs only once in any given address. Generally speaking, everything to the left of the @ delimiter is referred to as the local part of the address, usually a "mailbox" where a user reads his/her mail. A mailbox name often serves as a person's login name as well. In the example, jsmith is the local part, here signifying a mailbox as well as a login name. DNS is generally not involved in the local part of an address.
Everything to the right of the @ sign, on the other hand, is referred to as the "domain" name, which is based upon DNS and outlines where a mailbox is located. The complete domain name, also called the Fully Qualified Domain Name, in the example is uxa.cso.uiuc.edu. It consists of a sequence of symbolic labels, called subdomains, separated by periods.
In most Internet addresses, as in this example, the first subdomain (in listing from left to right) refers to an actual computer, or host, on which the mailbox resides. Uxa is the name of the host where the mailbox for jsmith is found. The next subdomain, cso, refers to the Computing and Communications Services Office (CCSO), the local organization which owns or runs the host uxa. CSO categorically falls under the next listed subdomain, uiuc, which denotes the University of Illinois. The subdomain listed farthest to the right in a complete domain name is called the top-level-domain. In this example, edu is the top-level-domain signifying the broad category of educational institutions.
It should be apparent that subdomains are listed in order from most specific to least, with each subdomain falling "inside" the one the the right. The general e-mail address with DNS format is as follows:
localpart@domain , often seen as the more specific form user@host.subdomain.subdomain.top-level-domainTop-level domains are the broadest category of domain names, and in the U.S., "generic" top-level domains denote the type of Internet site an address corresponds to: edu (educational institutions), com (commercial), gov (U.S. government), and mil (U.S. military), among others. Worldwide, top-level domain names are two letter abbreviations called "country codes" which specify the national affiliation of an address: au (Australia), fr (France), se (Sweden), uk (United Kingdom), etc.
Some popular addressing conventions resemble DNS but don't strictly conform to the system's model. One of these variations is the use of "pseudo"-top-level-domain names, which indicate a host computer's association with a wide-area, non-Internet network. Though accepted by many machines' mailer programs, these names are considered invalid by the Internet's formal DNS rules. Prominent examples of pseudo-top-level-domain names are bitnet, uucp and hepnet.
An address with one of these pseudo-top-level-domains will typically only have one other subdomain listed (to the right of the @ sign), and that subdomain name will usually specify a certain host belonging to the designated network. Thus the general address format would be user@host.pseudo-top-level-domain and an example of this type of address would be cgedney@uiucvmd.bitnet. A local mailer environment that is friendly toward pseudo-top-level-domain names in an address would have to intercept such messages and perform preliminary routing accordingly.
Another widely used variation of DNS is related to the abridging of the complete, or Fully Qualified Domain Names (i.e. user@host.subdomain.subdomain.top-level-domain) if the sender and recipient belong to the same subdomain. For example, the most abridged form of an address consists solely of the mailbox name, with no @ sign or other delimiters and subsequent domain names involved. This form may usually be used if the sender and recipient mailboxes are located on the same host computer. This means that the remainder of their domain names would be exactly equivalent, and mailer systems provide for this typing shortcut. When sending e-mail anywhere outside of such a very local mailing environment (with the exception of some users' ability to create mail ""aliases" or "nicknames"), it is strongly encouraged to use "the Fully Qualified Domain Name.
For example, many gateways exist at different geographical points between the Internet and BITNET and between the Internet and UUCP-based networks like Usenet.Gateways' role in linking networks relates to another significant variation of e-mail addresses that incorporates DNS. In many cases, "intelligent" mailer software will automatically forward messages being sent between different networks to an appropriate gateway. This would occur in a "transparent" manner, that is, without the user who is sending the message knowing anything about the gateway involvement.
For example, a message sent from a UIUCnet user on the CCSO Unix machine, uxh, to the user doug at a remote BITNET site called vmaway, would probably rely on an address of the form doug@vmaway.bitnet. The local mailer on uxh would transparently rewrite the address such that the message is sent first to UIUCnet's local BITNET/Internet gateway (CCSO's IBM mainframe vmd), which would in turn route the message to the remote BITNET host vmaway.Other times, mail must be explicitly sent through one or more gateways, requiring the user to specifically invoke them in the e-mail address. These addresses appear in the general format of mailbox%domain0%domain1@domain2 , where domain0 refers to the host where the mailbox is found, and domain2 and domain1 represent the sequence of gateways needed to get there, respectively.
In such an address it may be unclear as to how network machines interpret and evaluate the names and symbols in proper order such that the message arrives at its correct destination. An example would probably clarify this. As complex as it may appear, it is not difficult to determine the routing order of a message addressed to maryk%greco.hepnet%dove.uucp@blintz.bio.uiuc.edu.
Since the singular @ sign has highest precedence, the portion of the e-mail address to its right would be evaluated first. Thus, the message would initially be routed to the the host blintz of the bio subdomain. Blintz, serving as one gateway, would in turn examine the still remaining part of the address, interpreting the rightmost % sign as a secondary @ sign. This figuratively produces maryk%greco.hepnet@dove.uucp. The second routing task, then, would focus on sending the message on to dove of the uucp pseudo-domain. Dove, the name of another gateway, would examine the as yet "unconsumed" address portion, where the one remaining % sign would be changed into an @ sign, resulting in maryk@greco.hepnet. The mail would then be forwarded to the host greco (of the pseudo-domain hepnet), home of the mailbox maryk.
The @-% form of addressing will be seen less and less as strict DNS-based addressing becomes more universal on other networks besides the Internet.
Of course, several other e-mail address formats are encountered besides those derived from DNS. These addresses will use different syntaxes and have different punctuation marks as delimiters between fields. Such addresses will vary as networks with different e-mail systems vary. Not all types can be covered by the scope of this user guide, but since the greatest amount of e-mail traffic in UIUCnet flows between Internet, BITNET and UUCP-based sites, these addressing conventions will be discussed in the following sections.
A brief review of the characteristics and capabilities of each of these wide-area networks, along with a description of addressing schemes for internal e-mail messages will be presented first.
Sections 4-8 discuss how those address methods will vary for outside mail.
INTERNET
The Internet refers to a large collection of national and international networks that are based on the Transmission Control
Protocol/Internet Protocol (TCP/IP) protocol suite and which share a common addressing scheme based on the Domain Name
System (DNS). Sites on the Internet are currently linked over long distances with 56kbps, T1 (1.54 Mbps), or T3 (45 Mbps) lines,
with links on the local levels approaching 80Mbps or more in signalling speed. Network services offered on the Internet include e-mail exchange, file transfer, remote login and bulletin board news exchange (e.g. Usenet).
The Internet internal mail addressing scheme uses the Domain Name System (DNS), with a general address format of localpart@domain, often of the more specific form, user@host.domain where domain consists of one or more subdomains and a top-level-domain, separated by periods.
For example, of an address for internal Internet mail might be carolyn@uxh.cso.uiuc.edu.
BITNET's internal mailing scheme relies on every BITNET node having a unique name. Sites have "routing tables" that are based on these names. The tables are used for the forwarding of e-mail messages to their correct BITNET destinations. BITNET's internal mail uses an uncomplicated address syntax that also relies on these specific node names. Depending on the mail program being used, this internal address format will appear as something similar to user at node or user@node.
For example, cgedney@uiucvmd might be used.
In its most basic form, internal mail of a UUCP-based network is addressed according to the routing path from source host to destination host. The sender specifies an address consisting of a sequential list of host machines through which to send the mail, followed by the login name of the recipient. Each item is separated by a "!" sign. Thus, the general format is:
host1!host2!host3!...endhost!userwhere host1 is the source of the message and endhost is the destination host where the recipient's mailbox (user) is located. Host2, host3,... are the machines in between.
An example of the address of a message from a UUCP user named sbaker to his friend jcook at another UUCP site might look something like: pie!chili!pizza!jcook . To respond, jcook would reverse the order of the named hosts in the path, as in the address: pizza!chili!pie!sbaker.
This system can get complicated since a UUCP user's e-mail address depends upon who sends him/her a message. In practice, a few things help ease this situation. One is that many UUCP machines now have intelligent mailer programs that perform most of the routing task themselves. For users of these machines, it often is sufficient to provide only the name of the recipient's destination UUCP machine, as well as his/her user name. The sender's UUCP machine will take care of the rest. This address syntax would appear something as endhost!user . Some UUCP hosts support domain-like mailers, and their users may use an internal mail address of the form user@endhost.uucp.
Alternatively, many people advertise their UUCP addresses relative to some well-known point (host). That leaves the sender of mail with the task of getting the mail message to that known point; from there, he/she just copies the address that the other person has provided. Such a practice is best for users at un-registered UUCP sites, which are unknown to the rest of the UUCP network and thus can't be reached through automatic routing methods.
One reason for complication in e-mail addresses between networks is related to formal host names as cited within e-mail addresses. Often machines, or hosts, belonging to a certain wide-area network (such as BITNET, the Internet or UUCP-based networks) have officially registered names for identification and routing purposes. Some machines, especially those serving as gateways, belong to more than one network, and their proper host names, as used in an e-mail address, may depend on which network's addressing convention is being invoked. In other words, when using the address scheme of one network, the host will be called one name, but if using the addressing convention of a another network (to which the same host also belongs), it may be referred to by a different name.
For example, CCSO's IBM mainframe serves as the UIUCnet local gateway between BITNET and the Internet. In the internal BITNET address format, it is referred to by its designated BITNET node name of uiucvmd (as in the example address cgedney at uiucvmd). Since the machine is also part of the Internet, within the Internet's DNS-based e-mail syntax, it is cited as vmd within its complete domain name of vmd.cso.uiuc.edu (for example, cgedney@vmd.cso.uiuc.edu).The same is true for another CCSO Unix machine which serves as the local UIUCnet gateway between the Internet and UUCP-based connections. In a conventional DNS-based Internet address, this host is referred to as uxc within its complete domain name of uxc.cso.uiuc.edu (for example jsmith@uxc.cso.uiuc.edu) but in a UUCP-style e-mail address, it is invoked in its officially registered UUCP name of uiucuxc (for example, uiucuxc!jsmith).
This aspect of differentiating between multiple names for the same machine is especially relevant when dealing with "pseudo"-top-level-domain names (to be referred to from here on as pseudo-domain names) within an address. Recall that "pseudo"-domain names in an e-mail address indicate association with a wide-area, non-Internet network. Some examples of pseudo-domain names are uucp and bitnet.
Although the use of these names in an e-mail address is accepted by many local mailer systems, they are considered invalid on the Internet at large. Pseudo-domains appear to be based in DNS but don't really conform to DNS rules. They will appear in the general form of:
user@host.pseudo-domainwhere host is specified by its official name within the network cited by the pseudo-domain name (as opposed to an Internet name for that host).
For example, in the address cgedney@uiucvmd.bitnet, bitnet is the pseudo-domain name, and uiucvmd is the formal name of a node on BITNET. As mentioned before, within an Internet DNS-based address, that same host would have been cited as vmd.cso.uiuc.edu.Return to ContentsIn another example, address doug@uiucuxc.uucp, uucp is the pseudo-domain name and uiucuxc is an officially registered UUCP host name. If a mailer does accept addresses based on pseudo-domains, it means that some preliminary routing will have to be performed on the local level.
This usually consists of the mailer program transparently rewriting the address such that the message is first sent to an appropriate gateway, which will in turn forward the message on to its appropriate destination.
For example the address joe at vmusa should be written joe@vmusa.bitnet before being sent out from an Internet host.If the sender's Internet host doesn't have a modern mailer program installed, that style of address won't be accepted. In such a situation, the message would have to first be explicitly sent to the nearest Internet/BITNET gateway, which could then forward it on to the appropriate BITNET site. This is done by making the address of the previously explained (in Section 2) form:(In a more locally pertinent example, mail can be sent to people on campus that use the AISS PROFS e-mail system, which resides on a BITNET node machine named UICVMC. One may use an address of the form user@uicvmc.bitnet, where user is the person's PROFS "id.")
mailbox%domain1@domain2where gateway.domain is DNS-based and may consist of subdomains and a top-level-domain, and node is the formal BITNET node (host) name.or more specifically:
user%node.bitnet@gateway.domain
In the UIUCnet community, that Internet/BITNET gateway is CCSO's IBM machine, vmd, so the general form would be user%node.bitnet@vmd.cso.uiuc.edu. In a continuation of an example noted above, if a sender's Internet machine in the UIUCnet community couldn't resolve the address joe@vmusa.bitnet, the sender should send the message to the local gateway, vmd, though an address as: joe%vmusa.bitnet@vmd.cso.uiuc.edu.
Explicitly sending the message to the nearest UUCP/Internet gateway might be necessary, meaning an address of the form gateway!endhost!user where gateway is cited by its official UUCP name, or user%endhost.uucp@gateway.domain, where gateway.domain is DNS-based. For some mailer systems, a format appearing as endhost!user@gateway.domain will also work. For these three formats as well, routing to endhost must be registered and thus known to the gateway.
If mail is still rejected, it may mean the destination host machine is an un-registered UUCP site. In such a case, the address must specify a UUCP-style path to the destination machine, at least relative to a well-known (registered) UUCP site. In the UUCP world, such a well-known site is uunet. Most, if not all UUCP sites (including UIUCnet's local Internet/UUCP gateway) can correctly and automatically route mail with an destination address that begins with uunet!hosta!hostb!...user. This also means that if a provided return address contains many "hops," the address can be shortened to the right-most well-known site.
For example, an address like turkey!gobbler!uunet!potatoe!peas!user could probably be shortened to uunet!potatoe!peas!user and used successfully. This will carry over into potential addresses of the form:Return to Contentsuser%endhost%....host2%host1.uucp@gateway.domainwhere in either case host1 is a "well-known" UUCP site like uunet, and gateway.domain is DNS-based.
host1!host2!...endhost!user@gateway.domain
For other BITNET nodes which have simple mailers and that may have difficulty in getting Internet-destined mail messages routed correctly, there exists a VM/CMS "exec" program to help facilitate this task.
This program is available via the anonymous FTP utility from CCSO's Unix machine uxc.cso.uiuc.edu, from net/uiucnet/inetmail.exec.
For example, if someone on a UUCP machine wants to reach someone on UIUCnet whose internal address is carolyn@uxh.cso.uiuc.edu (and whose local Internet/UUCP gateway is uiucuxc), the address should be converted to something like:
host1!host2!...!uiucuxc!uxh.cso.uiuc.edu!carolynor perhaps the abridged form:
uiucuxc!uxh.cso.uiuc.edu!carolynReturn to Contents
This document was written by Carolyn A. Gedney . The URL for this document is: http://www.uiuc.edu/ccso/pubs/all/300/ug300.html
Final Notes
Although lower-case letters were used to explain syntaxes and examples in this article, in general e-mail systems are
case-insensitive. Mixed and upper-case are usually acceptable as well. This guide to e-mail addressing is certainly not complete
but should suffice for most cases. Different mail systems have varying capabilities and may not always behave according to the
norm. If suggestions given in this article don''t work, feel free to call the CCSO Systems Consultants at 333-6133 for further assistance with e-mail addressing.