laser-lovers@uw-beaver (03/04/85)
From: mendelson.es@XEROX.ARPA
I've been scooped!! I was about to send the message that is presented
below the dotted line when Brian Reid's masterful message "PostScript
and Interpress: a comparison", appeared in my mail. So I'll preface my
original message with these remarks.
I want to compliment Brian on a piece of work of outstanding excellence.
I appreciate its objectivity. I agree almost completely with his
technical characterizations of the differences between the two
languages. I believe that that agreement will show up clearly in what I
wrote, but my characterization is not nearly so scholarly as his. I
hope it is as objective as his.
We do have some differences of opinion and of knowledge, but I do not
propose to address them here. We'll leave them for another day, after
both comparisons have been presented.
Jerry Mendelson
---------------------------
The following information was prepared by Jerry Mendelson, and is
submitted to Laser Lovers as a contribution to the general discussion of
the characteristics of the two printing languages, Interpress and
PostScript.
Let me state up front that I am a retired Xerox Engineering Fellow, and
that I am currently working as a consultant to Xerox. This report was
commissioned by Xerox, but these are my observations, not those of
Xerox. Xerox did not exercise technical control over the content, nor
editorial control over the presentation, of the following information.
I have tried to be as analytic, objective, and honest as I possibly
could, but I know far more about Interpress than I do about PostScript.
I welcome equally honest and objective responses to this discussion
containing corrections, additions, other perspectives, what have you. I
trust that these discussions will lead to well reasoned recommendations
on a standard for the description of documents for representation on
imaging devices.
INTRODUCTION
A comparison of Interpress and PostScript is perhaps most appropriately
begun by recounting the history of the development of the two languages.
The following additional material on the history of the developments of
Interpress and PostScript is based on my first hand knowledge of events
that have transpired within Xerox during the past ten years, plus direct
conversations with the key participants in the developments of these two
languages. It is factual and complete to the best of my knowledge.
Others might be able to provide additional information of which I am
unaware.
MORE ON HISTORY
A FORTH like graphics/printing language was developed by, among others,
John Warnock before coming to Xerox/PARC. After coming to Xerox, John
put together another similar language which he called JaM.
Long before JaM was developed another printing format (Press) had been
developed and placed in use at Xerox. It was largely the brainchild of
Bob Sproull who is now at Carnegie-Mellon. Variants of this effort run
most of the Xerox internal printers on the original 3 MBit/Second
Ethernet which is still operational within Xerox. A graphics research
program known as Cedar Graphics was also pursued at PARC during this
same time frame.
Finally, leveraging off of all of these earlier activities, what I
designate as Research Interpress was created as a combined effort of a
number of people at PARC under the management direction of Chuck
Geschke. The principal developers were Bob Sproull, Butler Lampson,
John Warnock, and Brian Reid, with significant participation from Bob
Ayers. (There may have been others whom I have unintentionally slighted
because I am simply unaware of their roles.) The resulting Interpress
language contains benefits which derive from all of those very talented
people.
During the course of these developments Xerox permitted publication of
some of the Cedar Graphics research effort in a paper authored by John
Warnock and Douglas Wyatt titled "A Device Independent Graphics Imaging
Model for Use with Raster Devices", in the July 1982 Computer Graphics
Volume 16, number 3, pp. 313-320. Xerox also permitted Dr. Leo Guibas,
a Stanford professor, to use some of the Interpress related research
work as course material for a course at Stanford.
I led the engineering effort that extracted a suitable subset of the
Research Interpress language, and pushed it through the corporate
standards activity and into the product line. The first subset of this
language was published internal to Xerox in a Xerox System Integraton
Standard (XSIS 048201), titled Interpress 82 Electronic Printing
Standard, dated January 1982. Subsequent backward compatible revisions
have been internally released. The current externally released version
of of Interpress is designated as Interpress Electronic Printing
Standard (XSIS 048404), Version 2.1, dated April 1984, and is contained
in the documentation set available from Xerox. The Printing
Instructions portion of Interpress were added to Research Interpress,
and included in the internally published Version 2.0 in June, 1983.
After Interpress 82 had been extracted, pushed through the corporate
standards process, and incorporated into products, but before
Interpress, Version 2.0 had been created, Chuck Geschke and John Warnock
left PARC and formed Adobe Systems. They developed PostScript by
extending a combination of the work that Warnock had done prior to his
work at Xerox, plus the material that Xerox had permitted to enter the
public domain through the publication of the above referenced paper,
plus the material that Xerox had released to the Stanford course. They
carefully avoided including any of the advanced features that Sproull,
Lampson, and Warnock had incorporated in Interpress but that Xerox had
not publicly released.
GENERAL OVERVIEW
Again, some caveats about this general overview. It is not presented as
fact, but rather as a set of observations as interpreted by one reviewer
who has more familiarity with Interpress than with PostScript. Another
reviewer might reach a quite different set of conclusions. Also, please
note that this discussion deals with attributes of the languages, and
does not address issues related to particular implementations of the
language.
With their thread of common history it is quite natural that Interpress
and PostScript turn out to be very similar languages. Their fundamental
concepts and structures are substantially the same. I can only
speculate about the causes for their differences, but it appears to me
that the following issues are fundamental:
1. The developers of the two languages had different perceptions
of the application environments in which they were intended to operate.
2. Adobe was precluded from pursuing some of the elegant Xerox
proprietary techniques that were not part of the public domain.
As a consequence Interpress appears to be richer in higher level
structure. PostScript appears to be richer in its more basic
structures. These issues are discussed below.
Application Environment Considerations
The languages appear to have distinctively different viewpoints of their
application environments, and of how they properly fit into those
environments. The designers of Interpress took the position that the
functions of creation and composition are properly the role of creation
devices. Printers should print, and should do so in a highly efficient
and productive manner. Interpress was designed to describe the results
of the creation process to printers in a printer independent fashion.
The language recognized that there were some things in the printing
domain that the printer could best deal with, and that the creator would
not necessarily have knowledge about. It, therefore, made provision for
the creator to describe how the document description should adapt itself
to the specifics of what it finds in the printer's environment, without
knowing a priori what that environment would be.
Interpress can be used with a single printer that is tightly coupled to
the creator source, but it is also designed so that it can operate in a
networked environment in which there may be a multiplicity of
printer/servers. In this latter environment the creator may be highly
decoupled from the destination printer. Further, in such an environment
the printer/server must at all times be in control of its own actions,
and an Interpress document may not be allowed to cause the execution of
operations that interfere with that control. Interpress is designed for
machine to machine communication with no human interaction. In fact,
Interpress is designed under the assumption that the document may be
printed on a wide variety of printers over an extended period of time,
i.e. stored over an extended period of time, later retrieved, and then
printed.
The designers of PostScript appear to have taken a different viewpoint.
PostScript appears to assume an environment in which there is not so
clear a separation between the creator and the printer. The creator
appears to be in much closer contact with a specific target printer, and
possesses much more information on the details of that printer's
environment. In fact, a PostScript master appears to have the ability
practically to take over complete control of the printer's resources,
e.g. open, close, read, and write files. PostScript enables an almost
interactive environment with a human creator in the loop. While the
language is printer independent, a specific PostScript master appears to
be more closely coupled to a target printer than does an Interpress
master. PostScript enables much greater print time program control to
the PostScript master, and relies on that master to perform many of the
actions that Interpress defers to the printer. Further, Adobe has
chosen to enrich the more basic aspects of their language to make it a
more general purpose composition language. An example of an ideal
application for PostScript would be that of a fully functioned Graphics
Composition station working in conjunction with a designated printer of
known characteristics. It would clearly be a better language for that
application than would Interpress.
Computing Load Allocation
The printing of any complex page creates a significant computing task.
Interpress's design tends to force this task to the creator side.
PostScript's design tends to push this task to the printer side. As a
result Interpress masters tend to enable a high printer throughput
capability. PostScript masters tend to impose higher computing loads
that lead to slower throughput capabilities at the printer. Note the
use of the word "tends" in those last two sentences. What I am trying
to say is that the normal use of the natural characteristics of the two
languages would lead to the suggested results. However, these are very
general observations about the inherent characteristics of the
languages. It is clearly possible to create a PostScript master that is
highly efficient, and that does not force heavy computation loads on the
printer. It is also clearly possible to create an Interpress master
that does force heavy computation loads on the printer.
IDENTITIES AND STRONG SIMILARITIES
The two languages are strongly similar in so many ways that an
exhaustive presentation is beyond the scope of this brief analysis.
Here are some major points of identity:
Both Interpress and PostScript are document description languages. They
use a language to describe how to construct the image they want printed
on the page.
They are device-independent. The description of the page image is in
terms of the image, not of the device that the image is to be created
on. It is up to the printer to interpret this description, and to
create an image that matches it to the best of the printer's capability
to do so.
Both languages contain two distinct parts, a general purpose programming
portion, and a special purpose image generating portion.
They both use a FORTH-like postfix notation language, i.e. a reverse
Polish notation in which the operands precede their associated operators
in the presentation sequence.
They both use a stack-oriented processing structure. Operands are
pushed into a stack when they are received. Operands are popped from
the stack when their associated operator is received. Operator
execution results are generally pushed back onto the stack.
They both transmit their document representations to the printer in byte
streams. These bytes streams are broken down into "tokens", explicitly
defined in the Interpress implementation, implicitly defined in the
PostScript implementation.
Both token streams can be generated on the generator side, and executed
on the printer side, in one pass implementations.
Both language operate on "typed" operands. Operand types are generally
equivalent in the two languages.
Both languages have a generalized array processing capability. Such an
array is a collection of objects, not necessarily of the same type.
Both languages provide means to access arrays by an integer index that
designates a specific array entry by its relative position in the array.
Arrays may also be organized by using key, value pairs in arbitrary,
but paired locations. A value is then located by designating its
associated key.
Both languages use a universal coordinate system as a reference
framework. Both languages are heavily dependent on the use of identical
forms of transformation matrices. These are matrices that represent
the transformation from a user coordinate system to the reference
framework coordinates, and thence to the printer coordinate system.
Both use transformations to rotate, scale, and position objects on the
page.
Both languages enable the construction of procedures that can be
repetitively invoked. Procedures can be invoked by mechanisms that save
and restore the printer environment so that their effects can be
isolated from the rest of the page.
Both languages use very similar imaging models. The model is one in
which the image is incrementally built starting with a blank page. A
page image element, e.g. a character, a pixel array, or a geometrically
defined graphic structure, is added to the page by the following
process. A "stencil" representing the object is obtained, scaled to
size, rotated for orientation, and positioned at a desired location on
the page. An opaque "ink" of any color is then "painted" through the
stencil, overwriting anything that is currently present on the page.
Colors do not mix, they overwrite.
Both languages maintain a set of imaging control parameters in an array.
Both provide operators that are used to change the state of these
parameters.
Both languages provide for formally structured representations of fonts.
These representations include font name and font metrics as well as font
shapes. Both languages enable the creation of fonts with arbitrary
character generation techniques, subject only to the constraint that the
generation process is describable within the language. PostScript
includes language mechanisms for the control and use of a font cache
which dynamically stores raster encoded character instances.
MAJOR DIFFERENCES
In spite of their common ancestry the two languages do contain some
major and significant differences. These are catalogued in the
following paragraphs. The general sequence of presentation is to
present the strengths of Interpress first, followed by those of
PostScript.
Interpress Strengths
Total Environment
Interpress has all of the essential qualities of a stand-alone language.
However, Interpress has been designed so that it also fits well into an
applications and network environment such as the total XNS environment.
PostScript is a purely stand-alone language. It makes little, if any,
explicit provision to deal with its environment. However, it does
contain a number of capabilities which would enable it to do so. Their
explicit use for this purpose is not described within the available
Adobe documentation.
Examples of the integrability of Interpress with its environment abound.
It contains a sequenceInsertFile function that enables the inclusion of
files accessible to the printer within its total environment. Such
files can contain Interpress masters, fragments of interpress masters,
or even printer dependent object code that represent previously
decomposed Interpress masters, e.g. forms. The printer can utilize the
full XNS path name so that an Interpress master can reach out into the
entire XNS universe that is attached to the printer network. Interpress
takes explicit steps to establish a universal name space and to provide
for a central registry mechanism for the distribution and control of
names in this space. This provides a means for establishing a uniform
environment for an extended family of distributed printers.
Interpress contains printer instructions (see below) that are merged
with those transmitted by the Printing Protocol that is used to invoke
the printing of a document. This enables the document to contain some
printing instructions that are document dependent, and the transmission
protocol to provide some printing instructions that are user dependent.
Interpress makes provision for the interrogation of the printer
environment from the master so that the master can adapt itself to that
environment.
Page Independence
Probably the most important difference between the two languages is the
fact that an Interpress document possesses a well-defined structure.
Among the attributes of this structure is that it guarantees page
independence. Page independence means that the language description of
each page is totally independent of that of any other page. Ensuring
page independence is critically important for a number of reasons. It
enables the decomposition and printing of documents in arbitrary page
order. (Many printers find it desirable to print last page first. Full
duplex printing may require unusual page printing sequences.) It
enables the creation of utility routines to manipulate Interpress
documents. Such routines can perform such tasks as creating a new
Interpress master by pulling together pieces from existing masters, or
creating two-up, head-to-toe, or signature masters from existing
masters. Because of page independence these operations can be executed
without any consideration of the Interpress representation of the
document. They need only parse the Interpress master to locate the page
breaks which are clearly delimited.
PostScript documents are specifically free form and unstructured. The
language enables page independence to be achieved through externally
imposed programming disciplines, but the language itself provides no
such guarantee. In fact, in the absence of a carefully controlled
programming discipline it is practically certain that PostScript defined
documents will not exhibit page independence.
Printer Instructions
Interpress contains an extensive set of Printing Instructions. These
instructions enable the master to control the actions of the printer,
e.g. to invoke two-sided printing or special finishing such as stapling.
They also provide information necessary for the effective use of the
printer within a multi-user environment, e.g. who printed the document,
what its name is, whom to charge for the printing, the provision of
passwords to control who can authorize the final printing, and so on.
Printing Instructions also enable the declaration of resources that the
document will require, e.g. the files that it will use, the fonts and
font sizes that it will use. Not only do Printing Instructions enable
the control of the printing environment, they also enable an up-front
determination of the ability of a given printer to print a document
and/or enable it to gather the resources it needs to do so in the most
efficient fashion. These might be used by the printer itself, or by a
printing environment manager/dispatcher who determines which of a number
of printers should receive the document for printing. PostScript makes
no explicit provision for any of these functions.
Priority Important
The availability of multiple colors with opaque printing quality means
that rules must be established for the printing of overlapping objects
of different colors. PostScript takes no explicit cognizance of this
fact. The unambiguous appearance of the final page can only be obtained
if one assumes that PostScript defined objects are printed in the
sequence in which they are generated.
Some printers may not find it convenient to print objects in the order
in which they are created within the master. Interpress contains an
imaging parameter, priorityImportant, that is used to designate when the
printing sequence must match the Interpress presentation sequence. When
priorityImportant is not true the printer is free to image objects in a
sequence of its own choosing. This can increase printer performance in
many cases.
Compactness
Interpress specifies a compact encoding for the transmission of a
document. This encoding reduces to one or two bytes the number of bytes
required to designate an Interpress operator. It also includes special
encoding notations for compact representations of sequences of literals.
These encoding techniques substantially reduce the number of bits
required to either store or transmit a document. Simple utility
routines are available for the bi-directional translation from this
compact format to full English language form ASCII character
representations suitable for human usage. PostScript only employs this
latter form of more verbose representation.
PostScript Strengths
File Handling
PostScript contains a much more extensive and more powerful set of file
handling capabilities than those provided within Interpress. This
reflects the viewpoint of the language that the creator is closely
coupled to a dedicated printer. It permits the master to take over
control of the printer's file resources at print time. The Interpress
philosophy of multiple printers, highly decoupled in space and time, and
possibly with different file handling techniqes precludes this
capability. The Interpress philosophy of not letting the document take
over the machine resources also precludes this capability.
General Purpose Programming Capability
PostScript provides a much richer set of general purpose processing
capabilities. It provides a programmer wth a rich set of arithmetic,
control, looping, string processing, conversion between object types,
signalling, etc., operations. Its use of full ASCII text representation
makes it convenient for human programmers to write programs in the
language. All of this is in keeping with the PostScript viewpoint
previously described. Use of these capabilities can impose heavy
processing loads on the printer.
Interpress provides a sufficiently rich set of general purpose
processing capabilities to meet the needs of the environment for which
it was designed. It has been explicitly constrained in these
capabilities so as to reduce the processing load on the printer, and, in
keeping with the Interpress philosophy, force the processing load to the
creator.
Procedure Calling
Both PostScript and Interpress provide procedure calling capabilities.
Both make provision for the establishment of a working space containing
local variables for the procedure. Both make provision for saving and
restoring the state of the imager control parameters so that the side
effects of procedures can be well controlled. Interpress very tightly
couples all of these effects in its procedure creation and calling
mechanisms. PostScript provides a more flexible, and in many cases more
efficient, capability by separating these effects.
Graphic Imaging Capabilities
PostScript has a much greater set of graphic image creating capabilities
than that of the currently released version of Interpress. These
include the ability to invoke circles and Bezier curves in either solid
or dashed representations. PostScript also includes a clipping region
capability, i.e. a defined curve whose interior defines the region in
which the current imaging is permitted to occur. PostScript contain a
richer set of line joining constructs, including mitering, bevelling,
and rounding.
The currently released version of Interpress only contains the ability
to generate straight line bounded graphic constructs with mitering and
rounding at line joins. Circles, conics, and Bezier curves are included
in Research Interpress, and are scheduled for later releases of the
published language. Thus, the two languages will ultimately be brought
into much closer alignment in their graphic imaging capabilitapabilities
now, and Interpress won't do so until later.
SUMMARY
In summary, Interpress was designed for use by a large variety of
document creation and document output devices, to be suitable as a
general purpose electronic printing standard, a language for printers
capable of high performance. It was not designed as a general purpose
composition language to be used directly by a programmer. The
Interpress model assumes that the user creates Interpress Masters via
any variety of document editing and composition systems, including
graphics composition languages. It is clearly meant to be a printing
standard with emphasis on an organization that presents a clear
separation between theprocesses that Xerox believes belong in the
creation domain, and processes that Xerox believes belong in the
printer's domain. Postscript, on the other hand, appears to have been
designed for use both as a programmers composition language and as a
language for printers in a tightly coupled stand-alone configuration.
It does not draw clear distinctions between creation domain processes
and printer domain processes. It is an excellent language for full
capability graphics composition applications.