This document is from 08/04/94. It is provided mainly for historical interest
(and because we don't yet have a newer document).

A set of requirements were solicited from the Carnegie Mellon campus community. Those requirements were then used in conjunction with a review of commercial systems and available technologies to come up with a set of goals.

An important goal of the Cyrus Project is to ensure the technologies chosen are or soon will be in widespread use outside of Carnegie Mellon. This is to prevent becoming an island which can expose Carnegie Mellon to long term maintenance costs. By choosing technologies which are in widespread use, many interfaces and tools can be imported leveraging off work done by others.

Currently, there are four major platforms which must be supported as clients by the Cyrus mail system: Unix/X11, Macintosh, Windows, and DOS. The costs of creating and maintaining clients for all these platforms with the current mail system has proven to be a major resources drain. Even keeping up with new operating system releases on these systems has been a major burden. With AMS any improvements to clients, servers and administrative tools have to be developed locally. There are few groups outside of Carnegie Mellon which are making improvements to AMS.

Producing a mail package for a particular platform is beginning to require more than just creating a mail interface on that platform. With application programming interface (API) standards beginning to proliferate, it will be important to keep up with the standards that are strong for each platform. Mail APIs are used to create mail-enabled applications. Back-ends for the popular APIs, such as MAPI (lead by Microsoft), VIM (lead by Lotus), and AOCE (by Apple), which work with the Cyrus system, will be needed.

The use of the current mail system was reviewed in determining the project goals. The current mail system supports over 9000 active mail readers. Well over half of that population reads mail from a Macintosh. In addition, use of the DOS/Windows platform has been increasing steadily. The number of users reading mail from Unix workstations has remained steady.

About a third of the population uses multiple mail reading platforms within a given week. Students will read mail using the mail interface on the machine they are doing classwork on. Often, students are required to do class work on different platforms for different classes. The users have gotten used to mail clients which will provide seamless support for their movement from platform to platform preserving information about messages which have been read and personal options which control the behavior of the clients. In order to meet the expectations of this group, client mobility has become a major goal of the new mail system.

AMS was very successful with its attempt to integrate electronic mail and bulletin boards (bboards). This feature has greatly increased the use of bboards on campus. While, many people still contest whether it was wise to integrate mail and bboards as seamlessly as AMS did, it is clear that in order to support the widespread and varied uses of bboards on campus today, the new mail system must support thousands of simultaneous accesses to bboards and be able to support the rich set of uses of the current bboard system. These include restricted access to bulletin boards and an extensible addressing scheme to allow simple and complex mail address rules for posting to bulletin boards.

The ability to give selected sets of users access to mail folders and bboards has been a popular aspect of the current mail system. AMS, though, did not make it easy to make personal mail folders available to other users. This proved to be a major limitation in the current system and produced a system where individuals and groups needed administrative intervention to set up private forums. The new system should allow users' private folders to be made easily available to the entire user population as well as select groups of individuals.

The importance of not sending passwords over the network in the clear is becoming increasingly obvious. There is already a growing movement to not allow any further network protocols to get through the standards process which allow clear text passwords. Kerberos is the authentication system in use today with the Computing Services distributed services. The new system must support Kerberos authentication.

Interchange of documents containing a variety of objects between different mail systems is becoming quite popular. The MIME standard is increasingly being accepted as the means of such interchange. AMS is already capable of handling MIME messages. MIME will help provide for a smooth transition from AMS to Cyrus. MIME support is a requirement of the Cyrus mail system.

AMS uses the Andrew File System (AFS) as its network protocol. This use of a file system as the network access protocol for mail has many drawbacks. AFS has not been implemented for Macintosh and DOS machines. To support them, banks of translator machines were set up with a special micro mail server (mms) to make the mail in AFS available to Macintoshes and PCs. Design limitations of AFS clients and the AMS interactions with AFS made it necessary to have more translator machines than actual file servers providing service for mail. Performance problems on the translator machines have been a continuous problem. In addition, when the filesystem semantics are changed the entire mail system may have to change. This is happening now as AFS is being prepared to shift over to DFS (from the Open Software Foundation).

Using a remote filesystem as the network protocol for a mail and bboard system leads to many scale and availability problems. It is important that network protocols designed for mail be used rather than one for file systems. Preferably, the mail protocol should be a simple protocol to make implementation of clients easier.

In short, the major goals of the Cyrus project are to introduce a mail system to the Carnegie Mellon campus which:

• Is scalable to tens of thousands of active mail readers;
  
• Uses widely accepted standards-based technologies;
  
• Can be independently installed, managed and set up for use in small departmental computing facilities;
  
• Provides comprehensive support for clients on all major platforms in use at Carnegie Mellon;
  
• Preserves user profile information across sessions and between different clients on different platforms;
  
• Supports thousands of simultaneous accesses and varied uses of bulletin boards;
  
• Supports a disconnected mode of operation for the mobile user;
  
• Allows for easy sharing of private folders with select individuals;
  
• Supports Kerberos authentication;
  
• Does not base the network protocol for access to the remote mail stores on the campus file system.