Secure language for active documents & email

[Mark S. Miller <vlad!mark>]

Return-Path: <xanadu!kanga.parc.xerox.com!saraswat>
Date: Mon, 4 Jun 90 10:01 PDT
From: Vijay Saraswat <xanadu!parc.xerox.com!saraswat>
Subject: [Patricia.Miller@xxxxxxxxxxxxxxx: Special Talk-Nathaniel S. Borenstein]
To: xanadu!parc.xerox.com!januscore
Included-Msgs: <9005301725.AA21535@xxxxxxxxxxxxxxxx>,
               The message of 30 May 90 10:23 PDT from Patricia.Miller@xxxxxxxxxxxxxxx

Spent a few hours on Saturday with DHD Warren in Bristol, and he asked
me whether we had really identified some good application
to help push the Janus design and implementation effort. In fact, he
sees the lack of path-breaking applications of logic programming ideas
a significant handicap to the development of the field. 

Various subsets of us have thought on and off about a really good
driving application for Janus. The major proposals I have seen have been
for a distributed hypertext system, or a distributed active mail
system, or some kind of a LaTex-with-constraints document processing
language. The first two seem solid possibilities worth exploring. We
should allocate some bandwidth to discussing these issues.

Here is a thought-provoking abstract of a talk by Nathaniel Borenstein.
(Nathaniel got his PhD from CMU in 1986 on some user-interface
studies for the Unix Bags system (email/bboard/cboard system)
he had designed. He is a really good systems developmer.) Will find out
more and keep people posted. 

Vijay
----------------------

Date: Wed, 30 May 90 10:23 PDT
From: Patricia.Miller@xxxxxxxxxxxxxxx
To: vas@xxxxxxxxxxxxxxxxx
Subject: Special Talk-Nathaniel S. Borenstein


MAGICMAIL:  A Secure and Portable Language for Enhanced Email Services

			   A talk by			

                     Nathaniel S. Borenstein
                           Bellcore

		   Tuesday, June 5, 3:30 - 5:00

			   Wean 5409

Conventional electronic mail systems share a common paradigm that views
mail as basically an invariant object to be viewed passively by the
recipient.  This is true not only of standard text-only mail systems,
but also of multimedia mail systems such as Andrew and Diamond.  The
contents of a mail message, whether it be text, graphics, sound, or
animation, basically has no interactive capabilities other than to be
viewed by the recipients.  (Complex objects such as animations or
hierarchical drawings can be viewed from multiple perspectives, but the
interface for changing perspectives is not a part of the message itself,
just as text can be scrolled even though the scrolling mechanism is
external to the message.)

A few exceptions to this paradigm, if generalized, point the way toward
a new kind of email service.  Experimental systems (such as Hogg's
imail), and some small and well-constrained features in production mail
systems (such as return-receipt requests or multiple-choice votes in
Andrew and other systems), offer mechanisms whereby mail can embed
within it some form of computation.  In this paradigm, the mail can
interact algorithmically with the recipient, instead of merely passively
displaying its contents, to achieve some purpose.

The promise of algorithmic, or "active" mail, is very great.  One can
imagine email-based services productively applied to such diverse
applications as scheduling tasks, document routing,
information-gathering, and royalty collection on copyrighted material. 
However, there are two significant obstacles to the use of active mail,
namely security and portability.  Active messages should be powerful
enough for the kind of tasks just mentioned, but should not be able to
destroy, corrupt, or steal a recipient's resources.  Less critical, but
also important, the algorithms embedded in active messages should be
expressed in an interface-independent way, so that they can, like email,
be almost ubiquitously availability in heterogeneous environments.

In this talk, I will present an overview of MAGICMAIL, a language I have
designed and implemented which promises to overcome these obstacles.  

Incidentally, it should be noted that the constraints on active
electronic mail messages are a superset of the constraints on active
documents.  Therefore, MAGICMAIL might also be considered in the more
general context of documents with embedded computation.  However,
email's special constraints -- notably in the area of security and
asynchronous execution -- are such that the ideal languages for active
messages and active documents may not necessarily be identical.