Network Working Group                                          Alan KatzRequest for Comments: 1003                                       USC/ISI                                                              March 1987        Issues in Defining an Equations Representation StandardStatus of This Memo    This memo is intended to identify and explore issues in defining a    standard for the exchange of mathematical equations.  No attempt is    made at a complete definition and more questions are asked than are    answered.  Questions about the user interface are only addressed to    the extent that they affect interchange issues.  Comments are    welcome.  Distribution of this memo is unlimited.I.  Introduction    Since the early days of the Arpanet, electronic mail has been in    wide use and many regard it as an essential tool.  Numerous mailing    lists and newsgroups have sprung up over the years, allowing large    numbers of people all over the world to participate remotely in    discussions on a variety of topics.  More recently, multimedia mail    systems have been developed which allow users to not only send and    receive text messages, but also those containing voice, bitmaps,    graphics, and other electronic media.    Most of us in the Internet community take electronic mail for    granted, but for the rest of the world, it is a brand new    capability.  Many are not convinced that electronic mail will be    useful for them and may also feel it is just an infinite time sink    (as we all know, this is actually true).  In particular, most    scientists (apart from computer scientists) do not yet use, or are    just beginning to use, electronic mail.    The current NSF supercomputer initiative may change this.  Its    primary purpose is to provide remote supercomputer access to a much    greater number of scientists across the country.  However, doing    this will involve the interconnection of many university-wide    networks to NSF supercomputer sites and therefore to the NSF    backbone network.  Thus, in the very near future we will have a    large number of scientists in the country suddenly able to    communicate via electronic mail.    Generally, text-only mail has sufficed up until now.  One can dream    of the day (not so far in the future) when everyone will have    bitmapped display workstations with multimedia mail systems, but we    can get by without it for now.  I believe, however, that the new NSF    user community will find one other capability almost essential in    making electronic mail useful to them, and that is the ability toKatz                                                            [Page 1]RFC 1003                                                      March 1987    include equations in messages.    A glance through any scientific journal will demonstrate the    importance of equations in scientific communication.  Indeed, papers    in some fields seem to contain more mathematics than English.  It is    hard to imagine that when people in these fields are connected into    an electronic mail community they will be satisfied with a mail    system which doesn't allow equations.  Indeed, with the advent of    the NSF's Experimental Research in Electronic Submission (EXPRESS)    project, scientists will begin submitting manuscripts and project    proposals directly through electronic mail and the ability to handle    equations will be essential.    Currently, there exists no standard for the representation of    equations.  In fact, there is not even agreement on what it is that    ought to be represented.  Users of particular equation systems (such    as LaTex or EQN) sometimes advocate just including source files of    that system in messages, but this may not be a good long-term    solution.  With the new NSF community coming on line in the near    future, I feel the time is now right to try to define a standard    which will meet the present and future needs of the user community.    Such a standard should allow the interchange of equations via    electronic mail as well as be compatible with as many existing    systems as possible.  It should be as general as possible, but still    efficiently represent those aspects of equations which are most    commonly used.  One point to be kept in mind is that most equations    typesetting is currently being done by secretaries and professional    typesetters who do not know what the equations mean, only what they    look like.  Although this is mainly a user interface consideration,    any proposed standard must not require the user to understand an    equation in order to type it in.  We are not interested here in    representing mathematics, only displayed equations.    In this memo, I will try to raise issues that will need to be    considered in defining such a standard and to get a handle on what    it is that needs to be represented.  Hopefully, this  will form the    basis of a discussion leading eventually to a definition.  Before    examining what it is that could be or should be represented in the    standard, we will first review the characteristics of some existing    systems.2.  Existing Systems    There currently exist many incompatible systems which can handle    equations to a certain extent. Most of these are extensions to text    formatting systems to allow the inclusion of equations.  As such,    general representation and standards considerations were not a major    concern when these systems were initially designed.  We will examine    the three main types of systems: Directive systems, Symbolic    Language systems, and Full Display systems.Katz                                                            [Page 2]RFC 1003                                                      March 1987    Some text editing facilities simply allow an expanded font set which    includes those symbols typically used in mathematics.  I do not    consider these systems as truly able to handle equations since much    of mathematics cannot be represented.  It takes more than the Greek    alphabet and an integral and square root symbol to make an equations    system.    Directive systems are those which represent equations and formating    information in terms of directives embedded in the text.  LaTex and    EQN are two examples.  LaTex is a more friendly version of Knuth's    Tex system, while EQN is a preprocessor for Troff, a document    preparation system available under Unix.    With these Directive systems, it is usually necessary to actually    print out the document to see what the equations and formatted text    will look like, although there are on-screen previewers which run on    workstations such as the Sun.  Directive systems have the advantage    that the source files are just text and can be edited with standard    text editors (such as Emacs) and transferred as text in standard    electronic messages (a big advantage considering existing mail    interconnectivity of the various user communities).  Also, it is    relatively easy to make global changes with the help of your    favorite text editor (for example, to change all Greek letter    alpha's to beta's or all integrals to summation signs in a document.    This is generally impossible with the other types of systems    described below).    The primary disadvantage of these systems is that writing an    equation corresponds to writing a portion of a computer program.    The equations are sometimes hard to read, generally hard to edit,    and one may make syntax errors which are hard to identify.  Also,    people who are not used to programming, and typesetters who do not    actually know what an equation means, only what it should look like,    find specifying an equation in this language very difficult and may    not be willing to put up with it.    Full Display Systems are those such as Xerox STAR and VIEWPOINT.    The user enters an equation using the keyboard and sees exactly that    equation displayed as it is typed.  At all times, what is displayed    is exactly how things will look when it is printed out.    Unfortunately, VIEWPOINT does not allow the user to place any symbol    anywhere on the page.  There are many things (such as putting dots    on indices) which are not possible.  For those things which are    implemented, it works rather nicely.    Hockney's Egg is a display system which was developed at the UCLA    Physics Department and runs on the IBM PC.  It has the advantage of    being able to put any character of any font anywhere on the screen,    thus allowing not only equations, but things like chemical diagrams.Katz                                                            [Page 3]RFC 1003                                                      March 1987    Interleaf's Workstation Publishing Software system is not strictly    speaking an equations system, but equations may be entered via a cut    and paste method.  At all times, what one sees is what will be    printed out and one may put any symbol anywhere on the page.  The    problem with this system is that one HAS TO put everything in a    certain place.  It sometimes takes an enormous amount of work to get    things to be positioned correctly and to look nice.    Generally, Full Display Systems are specific to a particular piece    of hardware and the internal representation of the equations is not    only hidden from the user, but is in many cases proprietary.    Symbolic Language systems, such as Macsyma and Reduce, also allow    the entry of equations.  These are in the form of program function    calls.  These are systems that actually know some mathematics.  One    can only enter the particular type of mathematics that the system    knows.    We next will look at what should be represented in an equations    system.  We will want a representation standard general enough to    allow (almost) anything which comes up to be represented, but does    not require vast amounts of storage.3.  What Could be Represented?    We will first examine what it is that could be represented.  At the    most primative level, one could simply store a bitmap of each    printed equation (expensive in terms of storage).  At the other end