Tech-invite3GPPspaceIETFspace
959493929190898887868584838281807978777675747372717069686766656463626160595857565554535251504948474645444342414039383736353433323130292827262524232221201918171615141312111009080706050403020100
in Index   Prev   Next

RFC 1168

Intermail and Commercial Mail Relay services

Pages: 18
Informational

ToP   noToC   RFC1168 - Page 1
Network Working Group                                         A. Westine
Request for Comments: 1168                                    A. DeSchon
                                                               J. Postel
                                                               C.E. Ward
                                                                 USC/ISI
                                                               July 1990

              INTERMAIL AND COMMERCIAL MAIL RELAY SERVICES


STATUS OF THIS MEMO

   This RFC discusses the history and evolution of the Intermail and
   Commercial mail systems.  The problems encountered in operating a
   store-and-forward mail relay between commercial systems such as
   Telemail, MCI Mail and Dialcom are also discussed. This RFC provides
   information for the Internet community, and does not specify any
   standard.  Distribution of this memo is unlimited.

INTRODUCTION

   The evolution of large electronic mail systems testifies to the
   increasing importance of electronic mail as a means of communication
   and coordination throughout the scientific research community.

   This paper is a summary of the development of, and a status report
   on, an experiment in protocol interoperation between mail systems of
   different design. USC/Information Sciences Institute (ISI) began work
   on this experiment in 1981 and over the years has provided an
   evolving demonstration service for users to exchange mail between the
   Internet and a few commercial mail systems.

   Recently other organizations have begun to provide similar services,
   demonstrating the ongoing need for interoperation of the Internet and
   the commercial mail systems.  We believe that ISI's pioneering work
   in this area has promoted this expansion of service.

   These systems include the Internet mail system, the US Sprint
   Telemail system, the MCI Mail system, and the Dialcom systems. All of
   the systems were designed to operate autonomously, with no convenient
   mechanism to allow users of one system to send electronic mail to
   users on another system.

   The Intermail and Commercial Mail Relay (CMR) services described in
   this paper were developed to provide a means for sending mail between
   the Internet and these commercial mail systems.
ToP   noToC   RFC1168 - Page 2
   The Internet is an interconnected system of networks using the SMTP
   mail protocol, which includes the ARPANET, MILNET, NSFNET, and about
   700 other networks; mail relays allow the exchange of mail with
   BITNET, CSNET, and the UUCP networks as well.  To the users, this
   Internet looks like one large mail system with at least 100,000
   computers and at least 400,000 users.  Figure 1 illustrates the path
   of a message sent by a user on one Internet host to a user on another
   Internet host.  For more details on the Internet and connected
   networks (see Appendix A).

   As commercial mail systems came into popular use, it became clear
   that a mail link between the Internet and the commercial mail systems
   was necessary (see Appendix B).  More and more commercial and
   research entities needed to communicate with the Internet research
   community, and many of these organizations (for one reason or
   another) were inappropriate candidates for Internet sites.  The
   Intermail and CMR services allow these groups to communicate with
   Internet users by purchasing electronic mail services from commercial
   companies.

INTERMAIL

   Intermail is an experimental mail forwarding system that allows users
   to send electronic mail across mail system boundaries. The use of
   Intermail is nearly transparent, in that users on each system are
   able to use their usual mail programs to prepare, send, and receive
   messages.  No modifications to any of the mail programs on any of the
   systems are required.  However, users must put some extra addressing
   information at the beginning of the body of their messages.

               <<< Figure 1 - Internet to Internet Mail >>>

   The earliest version of Intermail was developed in 1981, by Jon
   Postel, Danny Cohen, Lee Richardson, and Joel Goldberg [1]. It ran on
   the TOPS-20 operating system and was used to forward VLSI chip
   specifications for the MOSIS project between the ARPANET and the
   Telemail system.  The original addressing model used in this system
   was called "Source Route Forwarding".  It was developed to handle
   situations in which a message might travel multiple hops before
   reaching its destination.

   Later, in 1983, Annette DeSchon converted Intermail into a more
   general-purpose mail-forwarding system, supporting forwarding between
   the Internet mail system and three commercial mail systems: Telemail,
   MCI Mail, and Dialcom [3,4].
ToP   noToC   RFC1168 - Page 3
   As it became apparent that the level of generality of Source Route
   Forwarding was not needed, and as Intermail gained acceptance among
   users, an easier approach to addressing was developed.  The new
   addressing model is called "Simple Forwarding".  This form of
   addressing, like Source Route Forwarding, appears at the beginning of
   the text of each message.  It can be used to include various Internet
   mail header fields in addition to the standard "To" and "Cc" address
   fields.  This format also allows the use of special address formats,
   such as U.S. postal addresses and TELEX addresses, which are
   supported by the MCI Mail system.  The Intermail system performed
   partially automated error handling.  Error messages were created by
   the Intermail program and were then either approved or corrected by a
   human postmaster.

   Figure 2 illustrates the pathways between the user mailboxes in the
   commercial mail systems and the user mailboxes in the Internet via
   the Intermail accounts and program modules.  Figure 3 shows the
   Intermail processing in more detail.

              <<< Figure 2 - Commercial Mail to Intermail >>>

                  <<< Figure 3 - Intermail Processing >>>

COMMERCIAL MAIL RELAY

   In 1988, the Commercial Mail Relay (CMR) was developed to run on a
   dedicated UNIX system, replacing the TOPS-20-based Intermail system.
   The CMR is a store-and-forward mail link between the Internet and two
   commercial systems, Telemail and Dialcom. The only remaining
   forwarding performed by the TOPS-20 Intermail system is in support of
   the MCI Mail system.  (This is planned for conversion to the CMR.)
   The CMR supports relay-style addressing in the "Internet to
   commercial system" direction, as well as Simple Forwarding in both
   directions.  One advantage of relay-style addressing is that users
   from different commercial systems can appear on Internet mailing
   lists.  Another advantage is that the reply features of most Internet
   user applications can be used by Internet users to respond to mail
   that originated on a commercial system. Unfortunately, since we do
   not have access to the address-parsing software on the commercial
   systems, it is not possible for users of the commercial systems to
   enter addresses directly into the message header, and they must
   continue to use Simple Forwarding.

   The CMR supports automated error handling, which enables the system
   to provide faster turnaround on messages containing addressing
   errors, and requires much less intervention from a human postmaster.
ToP   noToC   RFC1168 - Page 4
DESCRIPTION OF THE CMR SYSTEM

   The Multi-channel Memo Distribution Facility (MMDF) is used as the
   system mail software because of its notion of separating the mail
   queue into separate channels [5].  This makes it easy to dedicate a
   channel/queue combination to each commercial system.  Internet mail
   comes in over the standard SMTP port, and the system parses the
   destination address, queuing the message in the proper outgoing
   queue.  A tag can be added to outgoing traffic so that replies can be
   made without any special processing at the destination site.

   The CMR uses a relay mailbox on each commercial system.  Commercial
   users send mail to this mailbox with a Simple Forwarding Header (SFH)
   at the head of their message text.  Each channel, in addition to
   sending outgoing mail into the commercial system, reads all messages
   in the relay mailbox and places them in a spool file in the CMR host
   computer.

   The processing of the spool file is performed by a single daemon. It
   parses each commercial system message header to find the sender and
   subject, then it searches for and processes the SFH.  The SFH
   contains the destination Internet addresses.  Figure 4a illustrates
   the path of mail from the Internet to the commercial sytems. Figure
   4b illustrates the path from the commercial systrems to the Internet.
   Note:  MCI Mail is not yet implemented.

   The CMR employs a simple accounting mechanism:  a shell script counts
   the number of times a string marker occurs in the MMDF logs.  At the
   end of the month, another script uses an "awk" program to total the
   number of messages sent and received with each commercial system. The
   Commercial Mail Relay is being developed by Craig E. Ward.  Ann
   Westine served as the Postmaster for both Intermail and the CMR until
   March 1989.  Currently, our Action Office serves as Postmaster.
   Questions may be sent to "Intermail-Request@ISI.EDU".

          <<< Figure 4a - The Internet to Commercial Systems >>>

          <<< Figure 4b - Commercial Systems to the Internet >>>

COMMERCIAL SYSTEMS SERVED

   The CMR provides mail relay service between the Internet and two
   commercial electronic mail systems:  the US Sprint Telemail system
   and the Dialcom system.  A CMR connection to MCI Mail is under
   development.  MCI Mail is currently served by the TOPS-20 Intermail
   system.  See Appendix C for recent traffic data.
ToP   noToC   RFC1168 - Page 5
   Telemail is an international commercial service.  Some of the
   Telemail systems served by the CMR include MAIL/USA, NASAMAIL/USA,
   and GSFC/USA.  Some government agencies, such as NASA and the
   Environmental Protection Agency, have dedicated Telemail systems.
   Companies also exist that purchase bulk services from Telemail and
   resell the service to individuals.  Omnet's Sciencenet is a very
   popular example of this type of service.

   Dialcom is a commercial service similar to Telemail in that it has
   facilities for allowing groups to purchase tailored services, and
   some government agencies (such as the National Science Foundation and
   the U.S.  Department of Agriculture) have special group-access plans.
   The IEEE Computer Society also has a dedicated group service, called
   IEEE Compmail, which is open to members of the IEEE Computer Society.

   MCI Mail is operated by MCI and is marketed to large companies as
   well as individual users.

   Specific examples of the users of Intermail and the CMR are as
   follows:

   1) Scientists in Oceanography, Astronomy, Geology, and Agriculture
   use Intermail and the CMR to communicate with colleagues.  Many of
   these scientists have accounts on "Sciencenet", which is actually
   part of a Telemail system administered by Omnet.

   (2) The IEEE Computer Society's publication editors use the Dialcom
   system "Compmail" to manage the papers being prepared for their
   numerous publications.  Many of the authors are in university
   departments with access to the Internet. Intermail and the CMR
   support a significant exchange of large messages containing
   manuscripts.

   (3)  NASA uses Telemail systems for its own work and has extensive
   exchanges through its own relay service, as well as Intermail and the
   CMR, for communicating with university scientists on the Internet.

   Other developments to interoperate between the Internet and
   Commercial mail systems are:

      (1)  The Merit gateway to Sprintmail and IEEE Compmail

      (2)  The CNRI gateway to MCI Mail

      (3)  The Ohio State University gateway to Compuserve, and,

      (4)  NASA-Ames gateway to Telemail
ToP   noToC   RFC1168 - Page 6
ACCEPTABLE USE POLICY FOR INTERMAIL AND THE CMR

   The Internet is composed of many networks sponsored by many
   organizations.  However, all the Internet's long-haul networks are
   provided by U.S.  government agencies.  Each of these agencies limits
   the use of the facilities it provides in some way.  In general, the
   statement by an agency about how its facilities may be used is called
   an "Acceptable Use Policy".

   The various agencies involved in the Internet are currently preparing
   their Acceptable Use Policy statements.  Most of these are in draft
   form and have not been released as official agency statements as yet.
   None of these policies are currently available as online documents.

   In the least restrictive case, all bona fide researchers and
   scholars, public and private, from the United States and foreign
   countries (unless denied access by national policy) are allowed
   access.

   The Intermail and Commercial Mail Relay (CMR) systems at ISI are
   resources provided by the Defense Advanced Research Projects Agency
   (DARPA) for computing and communication.  Use of these resources must
   be limited to DARPA-sponsored work or other approved government
   business (or must otherwise meet the acceptable use policy of DARPA),
   only.

   However, DARPA, as a member of the Federal Research Internet
   Coordinating Committee (FRICC), has agreed to the FRICC draft policy
   for communication networks, which provides in part that:  "The member
   agencies of the FRICC agree to carry all traffic that meets the
   Acceptable Use Policy of the originating member agency".

   Thus, e-mail messages (i.e., "traffic") that meet the Acceptable Use
   Policy of an agency and pass through some facility of that agency
   (i.e., "the originating member") on the way to Intermail or CMR are
   allowed.

   The current member agencies of the FRICC are DARPA, NSF, DOE, NASA,
   and NIH.

   BITNET and UUCP (and other) networks are interconnected to the
   Internet via mail relays.  It is the responsibility of the managers
   of these mail relays to ensure that the e-mail messages ("traffic")
   that enter the Internet via their mail relays meet the Acceptable Use
   Policy of the member agency providing the Internet access.
ToP   noToC   RFC1168 - Page 7
   In addition, we cannot allow CMR or Intermail to be used simply as a
   bridge between two commercial systems, even though CMR has this
   technical capability.  At least one end of the communication must be
   related to FRICC acceptable use.

DETAILS OF CMR SYSTEM USE

   The CMR host computer is Internet host INTERMAIL.ISI.EDU
   (128.9.2.203).  The users of the commercials system are required to
   know the proper gateways between the Internet and other networks such
   as BITNET, CSNET, or UUCP.  Users on networks interconnected to the
   Internet likewise need to know how to reach the Internet to send mail
   through INTERMAIL.ISI.EDU to a commercial system.

   The relay connection to Telemail is through their host TELEMAIL/USA.
   The general syntax for Telemail addresses is
   "[USER/ORGANIZATION]HOST/COUNTRY", making the full address for the
   relay mailbox:

                      [INTERMAIL/USCISI]TELEMAIL/USA

   Users across the entire Telemail service can send mail to this
   address.  Users on the TELEMAIL host need only send to INTERMAIL.

   Internet users can use the basic Telemail format, append a
   "%TELEMAIL" to it, and mail to the resulting address as if it really
   existed on INTERMAIL.ISI.EDU, e.g.:

           [CWARD/USCISI]TELEMAIL/USA%TELEMAIL@INTERMAIL.ISI.EDU

   Note that the CMR system will accept anything before the "%TELEMAIL",
   that is, the CMR does not validate Telemail addresses before
   transmitting them to Telemail.

   The CMR handles Dialcom mail delivery in a similar way, but this
   system has what might be called "virtual hosts".  Groups can be set
   up with an alias system to allow easier intra-group access.  For
   example, both NSF and USDA share the same Dialcom host (157); but,
   while both groups send relay messages to Intermail, their actual
   fully qualified Dialcom mailboxes are different. For example, NSF's
   mailbox is NSF153, and USDA's mailbox is AGS9999.

   Mail going in either direction may use an embedded Simple Forwarding
   Header.  An SFH must be the first part of the message text.  It
   starts with a "Forward:"  field followed by a "To:" field.  "Cc:",
   "Subject:", and other fields may follow the "To:" fields. The SFH is
   terminated by a blank line.
ToP   noToC   RFC1168 - Page 8
   This is a template of an SFH:

      Forward: Destination-Network
      To: User@host1, User@host2,
           User2@host2
      Cc: User@host1
      Subject: This subject supercedes the subject in the host net header
      <Blank-Line>

   Dialcom syntax is "Host-ID:User-ID", for example, 134:ABC1234.  This
   format will work from any Dialcom host; but users in the same group
   as ABC would be able to use the user name, for example, JSMITH.

   Using the SFH format, mail to a Dialcom system could be sent as
   follows:

      To: Intermail@ISI.EDU
      Subject:  Test Message

      Forward: Compmail
      To: 134:ABC1234

      Here is the text of the message.

   Proper destination network names include ARPA, Telemail, Compmail,
   NSF-Mail, and USDA-Mail.

   It is possible for a user to make mistakes at many points in the
   process. Errors are handled as automatically as possible by the CMR.
   Many errors are caught in the standard Internet mail traffic, and
   users receive the usual error messages from the system.  Messages
   with incorrect commercial system addresses or faulty SFHs are also
   automatically returned to sender.  Messages that the software cannot
   handle are sent to the CMR's user-service mailbox, Intermail-
   Request@ISI.EDU.  This mailbox has been set up to take care of user
   problems and to be a central distribution point for user
   instructions.

PROBLEMS

   Several problems arise from the store-and-forward nature of the CMR.
   One of the biggest is that almost all of the commercial systems lack
   a machine-to-machine interface -- the CMR software must mimic a human
   user of the commercial system.  Another problem is that the Internet
   and a commercial system have different forms (or syntax) for
   electronic mail addresses.  A major goal of the CMR project is to
   make the link between networks as transparent as possible, allowing
   Internet users to use off-the-shelf mail programs.  Making commercial
ToP   noToC   RFC1168 - Page 9
   address formats fit the Internet standard is a major task [2].

   Compatibility with Internet addressing standards is also a concern.
   The commercial accounts are not able to take advantage of the
   transparency features of the Domain Name System (DNS) (see Appendix
   D); and some commercial addresses are incompatible with the Internet
   syntax--this requires Internet users to continue using the older
   methods.

   Another general problem to be solved is to reduce the amount of time
   needed to maintain the system.  Because most commercial systems force
   our software to mimic a human user, automatic error detection and
   handling are quite complex. The Intermail system requires human
   intervention in processing failed mail.  A goal of the CMR is to
   fully automate these processes.

   A related problem facing the CMR, as well as its predecessor
   Intermail, is the frequency with which commercial systems change
   their software.  The changes are usually minor and do not bother most
   human users; however, the CMR depends on being able to recognize
   certain strings.  To avoid the necessity of rebuilding the whole CMR
   when these strings change, most of the string markers are stored in
   ASCII files that are read at run time.

   The translation of commercial system addresses has created a new set
   of problems,  most of which are caused by the use of "special"
   characters by the commercial systems.

   Telemail uses square brackets ("[" and "]") around user names. While
   these characters are not special by Internet standards when found in
   the local part of an address, many (perhaps most) Internet mailers
   refuse to accept these characters unless they are quoted.  MMDF was
   modified locally to correct this.

   The square bracket problem is even worse for users of IBM mainframe
   machines, many of which are used on BITNET.  The square bracket is
   not a printable character on many BITNET IBM hosts, and all kinds of
   strange addresses can result from its use.

   The colon is another example.  Dialcom uses it as the delimiter
   between host and mailbox.  However, the colon is a special character
   in the Internet mail standard [2].  Users can avoid this problem by
   using the SFH and placing the Dialcom address at the beginning of the
   message text.  Although the CMR can accept addresses with colons,
   many Internet hosts and relays are unable to accept addresses that
   contain colons.  Mail with colons in the address fields is often
   rejected by Internet hosts and is returned to the Intermail-Request
   mailbox for error processing.  This can cause significant delays.
ToP   noToC   RFC1168 - Page 10
   Problems have also been caused by confusion about which hosts are
   mail relays between the Internet and other systems compatible with
   the Internet mail standard [2]. (e.g., BITNET, UUCP, and CSNET).
   When the CMR was implemented, a decision was made that the CMR would
   not keep track of these mail relays.  When a relay is changed, as the
   BITNET mail relays were in 1988, mail may be rejected because the
   host either no longer exists or refuses the mail.

   The mail relay problem is a subset of the larger problem of
   communicating information about new features and changes to the user
   community. Virtually none of the users of the CMR are local.  Many
   are hidden behind the veil of the commercial system.  (Dealing with
   commercial system customer support people has proven to be
   frustrating -- few of them seem to understand the concept of
   machine-to-machine exchanges.) Enhancements to commercial software
   that necessitate minor changes can disrupt some CMR users for days.

   Another problem that has not been adequately solved is validation of
   commercial system addresses and processing of failed commercial
   system mail.  The Telemail system will not validate a user/host
   combination until after the full text of the message has been
   transmitted.  If a long message is sent to an invalid address, it can
   be very expensive in terms of wasted time and connect charges.

   Telemail also gives inadequate information when the host is correct
   but the user name is not.  The failed mail notice received from
   Telemail is of little use to either a human reader or the CMR
   software.  The only information that Telemail returns is the message
   ID number -- it provides no subject, and no text to distinguish the
   message from the numerous others that pass through the mailbox.

   Dialcom does a better job of validating addresses.  If an address is
   not recognized, the system immediately prompts for a correction.  A
   simple <RETURN> will delete the invalid address from the list.

   The commercial systems are geared for paying customers to send and
   receive mail to other paying customers.  They are not equipped to
   handle reverse billing, or "collect calls."  ISI is currently charged
   for connect time needed to transmit and receive mail to and from
   other Internet sites.  A possible solution to this problem would be
   to extend the CMR. to include accounting and billing procedures that
   would pass the costs of CMR to its users.

   What had been GTE Telemail became Sprint SprintMail, Telenet became
   Sprintnet, and the host TELEMAIL/USA became SM66/USA.

   In April 1990, Sprint installed its X.400 implementation.  For the
   time being, the old-style Interconnect syntax will work. The CMR
ToP   noToC   RFC1168 - Page 11
   telemail channel and the Simple Forwarding Header (SFH) processor,
   were modified to accept either format in the SprintMail "From" field.

   Sprint uses the following syntax for X.400:

                      (O:USCISI,UN:INTERMAIL,TS:SM66)

   The SFH processor will "translate" this into:

                 /O=USCISI/UN=INTERMAIL/TS=SM66/%TELEMAIL

   The channel program will reverse the process.  In the translation,
   parentheses become slashes, colons become equal signs and commas
   become slashes and vice versa.

   Unfortunately, the translation algorithm is not foolproof.  A
   Sprint/Internet relay did not use the same field names and values as
   those in SprintMail.  Consequently, a CMR translated address can not
   be sent unmodified to Sprint's relay, Sprint.COM, and Sprint.COM
   processed addresses cannot be sent unmodified to the CMR.

   From experimentation, the modifications necessary to a CMR processed
   address to make it acceptable to Sprint.COM are (1) take the "non-
   standard" X.400 fields of "UN" and "TS" and prepend "DD." to them,
   (2) add the country field and code (C:US) and (3) add the Telemail
   administrative domain name (ADMD:Telemail).  The above example would
   become:

    /O=USCISI/DD.UN=INTERMAIL/DD.TS=SM66/ADMD=TELEMAIL/C=US/@Sprint.COM

   The country code must be changed from "US" to "USA."  The CMR queue
   name must also be appended: "%TELEMAIL."

   The situation is further complicated by Sprint's decision to only
   relay mail to and from its own administrative domain.  Other X.400
   ADMDs may be added in the future if payment problems can be overcome.

   SprintMail encoded Internet addresses are not parsed correctly by the
   SFH processor, but that should not be a major problem -- who on the
   Internet is going to send to the commercial side of the relay?

   When the NSF decided to terminate NSFMAIL, it became clear that the
   CMR Project needed a way to get news out to the commercial users.
   The CMR channel programs now are able to append a news file to the
   end of messages going into the commercial networks.  After
   transmitting a message, each channel checks for a news file with the
   channel name and if present, sends it.
ToP   noToC   RFC1168 - Page 12
   The biggest costs of the CMR are the connect times to the Sprintnet
   X.25 network and the commercial machines.  Making the CMR transmit
   faster is the current number one problem.

   Three strategies are being pursued:

      - Improve the implementation of the current method

      - Change the method to take advantage of changes in the commercial
        software

      - Upgrade the modems and increase the number of phone lines

   For a list of known problems or bugs in the CMR software, see the
   Appendix of the program logic manual [6].

FUTURE DIRECTIONS

   No software project is ever completed, and the CMR is no exception.
   There are many possible extensions, some more difficult than others.

   One addition that will be made to the CMR is a channel for
   interacting with MCI Mail.  MCI Mail is one of the original TOPS-20
   commercial systems that were serviced by Intermail; the CMR will need
   to replace this function before all of the TOPS-20 machines are
   removed from service on the Internet.

   The adaptability of the CMR is such that adding new commercial
   systems should not be a major problem.  Additional commercial systems
   under consideration include General Electric's GENIE, Western Union's
   EasyLink, and Compuserve.

   One possible addition to the CMR system could be maintenance of a
   list of gateways.  This would allow commercial system users to
   incorporate the native address formats of other networks into the
   SFHs.  An advantage of this would be that users could simply tell the
   CMR to forward a message to BITNET, for example, and the CMR would
   find the gateway and properly format the address for that gateway.

   To increase the ease of use to Internet users, the system might treat
   each commercial system as an Internet host and create DNS database
   records for them.  This would allow users to send mail to a non-
   Internet user at an Internet-style domain name.

   Another improvement would be the possibility of accepting X.400-style
   addressing. The current system rejects them.
ToP   noToC   RFC1168 - Page 13
   In order to further reduce the hazards of string changes in the
   commercial system software, an AI component could be added to the
   commercial system interfaces.  Such an AI component might be able to
   "figure out" what marker a new prompt represents and to remember it.

ACCESS INFORMATION

   For instructions on how to use Intermail and CMR contact Intermail-
   Request@ISI.EDU.

REFERENCES

   [1]  Cohen, D., "A Suggestion for Internet Message Forwarding for
        MOSIS", IEN-180, USC/Information Sciences Institute, March 1981.

   [2]  Crocker, D., "Standard for the Format of ARPA Internet Text
        Messages", RFC-822, University of Delaware, August 1982.

   [3]  DeSchon, A. L., "MCI Mail/ARPA Mail Forwarding", USC/Information
        Sciences Institute, ISI Research Report, RR-84-141, August 1984.

   [4]  DeSchon, A. L., "INTERMAIL, An Experimental Mail Forwarding
        System", USC/Information Sciences Institute, ISI Research
        Report, RR-85-158, September 1985.

   [5]  Kingston, D., "MMDF II: A Technical Review", Usenix Conference,
        Salt Lake City, August 1984.

   [6]  Ward, C. E., "The Commercial Mail Relay Project:  Intermail on
        UNIX", USC/Information Sciences Institute, 1988.
ToP   noToC   RFC1168 - Page 14
APPENDIX A

   The Internet and Connected Networks

   The Internet is a network of networks interconnected by gateways or
   routers. The common element is the TCP/IP protocol suite.  The
   Internet now includes approximately 800 networks and 100,000 host
   computers.  The Internet is made up of local area networks in
   research institutes and university campuses, regional networks, and
   long-haul networks.  These resources are supported by the using
   organizations and by several US goverment agencies (including DARPA,
   NSF, NASA, DOE, and NIH).   The long-haul networks in the Internet
   are the ARPANET, the MILNET, the NSFNET Backbone, the NASA Science
   Internet (NSI), and the DOE Energy Science Network (ESNET).

   Other systems using TCP/IP or other protocols may be networks of
   networks or "internets" with a lower case "i".  The capital "I"
   Internet is the one described above.

   There are other networks with (semi-) compatible electronic mail
   systems. These include BITNET (and EARN and NETNORTH), UUCP (and
   EUNET), CSNET, ACSNET, and JANET.  Users of electronic mail may not
   necessarily be aware of the boundaries between these systems and the
   Internet.

   The Domain Name System (DNS) is a mechanism used in the Internet for
   translating names of host computers into addresses.  The DNS also
   allows host computers not directly on the Internet to have registered
   names in the same style.

   BITNET (Because It's Time NETwork)

   BITNET has about 2,500 host computers, primarily at universities, in
   many countries.  It is managed by EDUCOM, which provides
   administrative support and information services.  There are three
   main constituents of the network: BITNET in the United States and
   Mexico, NETNORTH in Canada, and EARN in Europe.  There are also
   AsiaNet, in Japan, and connections in South America.  Gateways exist
   between BITNET and the Internet.  The most common gateway used is
   CUNYVM.CUNY.EDU.

   CSNET (The Computer + Science Network)

   CSNET has 180 member organizations, primarily computer science
   research laboratories at universities and research institutes,
   including international affiliates in more than a dozen countries.
   CSNET has adopted DNS-style names for all its host computers.  It is
   administered by the University Corporation for Atmospheric Research
ToP   noToC   RFC1168 - Page 15
   (UCAR) and provides administrative support and information services
   via the CSNET Information Center (CIC) at Bolt Beranek and Newman
   (BBN). The gateway between CSNET and the Internet is RELAY.CS.NET.

   Note: CSNET and BITNET have officially merged into a single
   organization as of October 1, 1989.

   UUCP (UNIX to UNIX Copy)

   UUCP is a protocol, a set of files, and a set of commands for copying
   data files from one UNIX machine to another.  These procedures are
   widely used to implement a hop-by-hop electronic mail system.  This
   simple mechanism allows any UNIX host computer to join the system by
   arranging a connection (dial-up or permanent) with any UNIX host
   already in the system.  In the basic UUCP system, mail is source
   routed by the sending user through a path of connected hosts to the
   destination.  Currently, there are databases of connection
   information (UUCP maps) and programs (pathalias) that aid in
   determining routes.  There is some use of DNS-style names by UUCP
   hosts.  EUNET is a UUCP-based network in Europe, and JUNET is a
   similar net in Japan.  These international branches of UUCP use DNS-
   style names as well.  There are many hosts that may relay mail
   between UUCP and the Internet.  One prominent gateway is
   UUNET.UU.NET.

   JANET (Joint Academic NETwork)

   JANET is the primary academic network in the United Kingdom, linking
   about 1,000 computers at about 100 universities and research
   institutes.  JANET has a domain name system similar to that of the
   Internet, but the order of the domain name parts is opposite (with
   the top-level domain on the left).  The protocols used in JANET are
   the UK "Coloured Books".  The primary gateway between JANET and the
   Internet is NSFNET-RELAY.AC.UK.

   ACSNET (Australian Computer Science Network)

   ACSNET is the principal electronic mail system for the computer
   science and academic research community in Australia, connecting
   about 300 sites.  It works similarly to UUCP.  ACSNET has a domain
   naming syntax similar to that for Internet domains.  The gateways
   between ACSNET and the Internet are MUNNARI.OZ.AU and UUNET.UU.NET.

APPENDIX B

                         <<< Mail Systems Map >>>
ToP   noToC   RFC1168 - Page 16
APPENDIX C

   March 1990      sent    read    total

   Telemail        1291    768     2059
   MCI Mail        56      44      100
   Compmail        634     306     940
   NSF-Mail        370     243     613
   CGnet           171     82      253
   USDA Mail       6       1       7
   BellSouth       6       10      16
   Other           0       0       0

   Total           2534    1454    3988
   Days in Month   31
   Messages per Day        128.65


   April 1990      sent    read    total

   Telemail        1363    696     2059
   MCI Mail        40      39      79
   Compmail        771     329     1100
   CGnet           361     191     552
   USDA Mail       28      31      59
   BellSouth       98      73      17
   Other           0       0       0

   Total           2661    1361    4022
   Days in Month   30
   Messages per Day        134.07


   May 1990        sent    read    total

   Telemail        1007    561     1568
   MCI Mail        23      12      35
   Compmail        815     359     1174
   CGnet           406     210     616
   USDA Mail       12      5       17
   BellSouth       167     93      260
   Other           0       0       0

   Total           2430    1240    3670
   Days in Month   31
   Messages per Day        118.39
ToP   noToC   RFC1168 - Page 17
APPENDIX D

   The Domain Name System

   The Domain Name System (DNS) provides for the translation between
   host names and addresses.   Within the Internet, this means
   translating from a name, such as "ABC.ISI.EDU", to an IP address such
   as "128.9.0.123".  The DNS is a set of protocols and databases.  The
   protocols define the syntax and semantics for a query language to ask
   questions about information located by DNS-style names. The databases
   are distributed and replicated.  There is no dependence on a single
   central server, and each part of the database is provided in at least
   two servers.

   In addition to translating names to addresses for hosts that are in
   the Internet, the DNS provides for registering DNS-style names for
   other hosts reachable (via electronic mail) through gateways or mail
   relays.  The records for such name registration point to an Internet
   host (one with an IP address) that acts as a mail forwarder for the
   registered host.  For example, the Australian host "YARRA.OZ.AU" is
   registered in the DNS with a pointer to the mail relay
   "UUNET.UU.NET".  This gives electronic mail users a uniform mail
   addressing syntax and avoids making them aware of the underlying
   network boundaries.
ToP   noToC   RFC1168 - Page 18
SECURITY CONSIDERATIONS

   Security issues are not discussed in this memo.

AUTHORS' ADDRESSES

   Ann Westine
   USC/Information Sciences Instutite
   4676 Admiralty Way
   Marina del Rey, CA  90292-6695

   Phone:  (213) 822-1511
   EMail:  Westine@ISI.EDU

   Annette DeSchon
   USC/Information Sciences Instutite
   4676 Admiralty Way
   Marina del Rey, CA  90292-6695

   Phone:  (213) 822-1511
   EMail:  DeSchon@ISI.EDU

   Jon Postel
   USC/Information Sciences Instutite
   4676 Admiralty Way
   Marina del Rey, CA  90292-6695

   Phone:  (213) 822-1511
   EMail:  Postel@ISI.EDU

   Craig E. Ward
   USC/Information Sciences Instutite
   4676 Admiralty Way
   Marina del Rey, CA  90292-6695

   Phone:  (213) 822-1511
   EMail:  Ward@ISI.EDU