Network Working Group Karl Kelley
Request for Comments 184 University of Illinois
NIC 7128 6 July 1971
Proposed Graphic Display Modes
The ARPA Network node at the University of Illinois' Center for
Advanced Computation is somewhat different from other nodes in that
we are not simply attaching an existing computer center to the net.
We are in the process of establishing the computer system
specifically for use of the ILLIAC IV and the Network. In this mode
we are establishing operating systems, network interface and utility
routines, and ILLIAC IV routines to be used over the network.
In the field of computer graphics we are in the process of building a
system essentially from scratch. The building blocks of this
capability comprise a small -- but growing -- collection of display
hardware and a small cadre of persons with experience on separate and
unique graphics equipment at the University of Illinois. Starting as
we are with little-or-no system type software for computer graphics,
we have a once-only opportunity to provide the system with computer
graphics applications and utility programs which encompass all the
features and capabilities that have heretofore been available only in
bits and pieces at various separate installations.
It is apparent at the outset that the design for this system will be
heavily weighted toward a network-type usage. For this reason we are
eager to ensure that our system data structures, files, etc., be as
nearly compatible the Network Graphics Protocol as is practicable.
Our initial planning and first-version system will be pointed at the
network type of operation and we hope to stay flexible enough to
employ the Network Protocol on a local basis (between our PDP-11 and
the B6500) as the protocol is developed.
We have been considering (in the planning of our system and pondering
the protocol problem) just what display modes we would want to have
available and thus would want the protocol to include. The purpose
of this RFC is to outline our initial thoughts on the matter and to
interact with other nodes about how they can/should be included in
the protocol. We intend here not to belabor display modes which are
certain to be needed everywhere, such as vectors, points, and
characters, but rather to summarize those and outline in more detail
only those which are slightly different.
The display system, and the network protocol, will require something
like the following list of display types:
1. Points | Including normal points,
< plot a symbol at a point, plot
| a point with intensity
2. Lines(two-point) | These two (2 and 3) include
< visible and not visible, dotted,
3. Vectors (from present beam | dashed, overbright, and ?
location to a point) \
4. Character Streams
5. Viewport and Window
Specifications (ala LDS-1)
6. Transformations (scaling and
rotation) of Instances
7. Equipment-Specific Byte Streams
8. Read-Back of Keyboards,
Function Buttons, Cursors, Etc.
It seems clear that some type of list or ring structure will be
needed to handle our display files. Whether that structure need be a
part of the protocol is not evident (after all, you could just send
the equipment byte-stream), but it is our feeling that it will be
needed. It is evident that the protocol must anticipate the needs of
various popular display devices as CalComp, Computek, and similar
storage displays, interactive displays such as Adage, LDS-1, Vector
General, IDIOM, grey-level displays like the PEP-1 and raster-
oriented gadgets like the Gould, Versatec, and garden-variety line
The standard display element types given above are assumed to be in
common use. A point, for example, is a pen-down command followed by
a pen-up command on a plotter. On an interactive device it is merely
an intensification of the beam. To plot a symbol at a point the pen
(beam) is moved to the point with pen up (beam off) and then the
symbol is plotted incrementally with the pen down (beam on).
Graphics devices with beam intensity control will be expected to
handle the "plot-point-with-intensity" format.
In terms of the standard display types the only difference between
lines of the two-point form and vectors is that in the former, four
data items are needed for each segment while in the latter, only two
data items are needed. In either case, the user should be able to
reposition the pen (beam) absolutely by drawing an invisible line.
He should also be able to plot dotted or dashed lines without having
to separately specify each point or short vector. In instances where
emphasis is needed it is useful to be able to selectively intensify a
line, or in the case of interactive displays, make it blink.
Character streams are used either for text or for labeling drawings.
In most applications the character stream is specified by its
starting location in screen coordinates, the number of characters,
and the location of a buffer of characters to be used. In some
purely output graphics systems the character stream is specified via
a format similar to printer output, with the characters being placed
in the specified area on the display.
Items 5 and 6 of the above list primarily apply to displays like
Sketchpad and the Evans and Sutherland display. Viewport is taken to
be synonymous with "observer parameters in the object space" while a
window means "selected portion of the display surface".
Transformations of instances refers to a feature of Sketchpad which
allows multiple uses of the same "pattern" for a graphical element.
Because new equipment is constantly coming into use, and special-
purpose equipment is available at some nodes, it is prudent to have
available a capability for sending equipment-specific information
over the net as a part of the protocol. Such information would be in
the form of byte streams formatted according to the equipment
specifications and pointed at the proper node and equipment. It
would not be expected that each node be able to interpret the
nonstandard byte stream. Also very equipment specific is the
information passed from an interactive device back to the originating
program. Elements such as joystick or cursor position, lightpen
hits, function buttons pressed, etc., are inherently dependent upon
the device employed. Although these devices are widely used, their
general dependence upon display buffers, display lists, or interrupts
is of special concern to the network graphics protocol.
We are interested in expanding upon the uses of grey-scale display
modes for representation of computer-generated data and for three-
dimensional object representation. To facilitate this, our system
will have available at least four, program-callable display element
types for production of pictures on grey-level display devices. The
initial names for these modes are the procedure names: GRIDAREA,
MATRIXAREA, SCANLINE, and SCANPOINT. The following paragraphs will
outline how the modes operate both from a user and a data-
communications point of view. The specific hardware involved is not
specified, nor do we ignore the possibility that hardware can be
designed to operate this way directly. For example, the SCANLINE is
precisely the way one display does operate. In the interim, however,
software sits between these procedures and actual devices.
GRIDAREA The user specifies, in an initial call, the size
-------- grid he wishes to use and the number of intensity
levels he will need. Subsequent calls send, as
data items, the i, j locations of an area and a
byte for intensity.
The initializing call is GRIDSET(N, M, IRNGE)
|<--- I ------>|
where N is the number of spaces --- +--+--+--+--+--+--+--+--+
across, M is the number down, ^ |--|--|--|--|--|--|--|--|
and IRNGE tells how many grey | |--|--|--|--|--|--|--|--|
levels to use. This is primarily J |--|--|--|--|--|--|--|--|
for grey-scale displays or | |--|--|--|--|--|--|--|--|
pseudogrey-scale displays. v |--|--|--|--|--|--|--|--|
On a Gould Electrostatic Printer-Plotter, for example, the system
would expect to have dot patterns with which to fill areas in order
to simulate the greys. For purposes of other displays, a SETSQUARES
procedure should be available so that the user can specify various
kinds of cross- hatching and character filling to apply to the grid
To enter an item of data the procedure is
1 <= I <= N
GRIDAREA(I, J, LEVEL) 1 <= J <= M
1 <= LEVEL <= IRNGE
where I and J select an area on the grid, and LEVEL tells how to fill
it. Obviously, for this kind of display mode some provision must be
made to end the picture because the servicing routine will have to
work on it from either the top or the bottom in a sweep mode. A
procedure call of GRIDAREA(0, 0, 0) will terminate it. The GRIDAREA
mode is very similar to the following, MATRIXAREA, with the primary
difference being that the specification of areas is random and areas
which are not specified will be left blank.
MATRIXAREA The user specifies an area size and a "pseudo-
---------- character" set, then writes from left to right,
top to bottom, much like a line printer, using
bytes to specify which of his "pseudo-character"
set to use.
The initializing call is MATRIXSET(N, M, DEFINITION, CODE)
The display mode is raster ________________________
oriented, and each "pseudo- | _ _ _ _ _ _ |
character" will be on a N x M | |_| |_| |_| |_| |_| |_||
matrix of dots. (A later | _ _ _ |
embellishment for printers | |_| |_| |_| . . . |
would include matrix of characters.) | _ _ |
Parameters DEFINITION and CODE | |_| |_| . . . |
are both arrays, used together to | . \ \ |
specify the "pseudo-character" set. | . \ \ |
DEFINITION is packed with bits |________\___\___________|
according to the following scheme: \ \
\ / \
\/ ________ \
/ |_|_|_|_| \
/ |_|_|_|_| \
| |_|_|_|_| |
| |_|_|_|_| |
\ |_|_|_|_| /
\ |_|_|_|_| /
n - bits
|XX| | |XX|
s | |XX|XX| |
i |XX| | |XX|
b |--|--|--|-- +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+--\
| |XX|XX| | <=== |1|0|0|1|0|1|1|0|1|0|0|1|0|1|1|0|1|0|0|1|0|1|1|0|\
- |--|--|--|--| +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-\
m |XX| | |XX| \--v--/ \--v--/ \--v--/ \--v--/ \--v--/ \--v--/
|--|--|--|--| n n n n n n
| |XX|XX| |
The bit stream for the definition is irrespective of word boundaries.
We leave it up to the MATRIXSET routine to be able to undo this.
Obviously, for user convenience we have some standard sets they can
use to save having to define their own, and if they want to define
their own, we make routines to ease the pain.
The array CODE, on the other hand, is a byte-stream, i.e., a stream
of eight-bit groups of bits which will correspond to each of the
groups of (N x M)-bits. This allows 256 pseudocharacters for one
To enter an item of data in this mode the procedure is
ARRAY is a buffer location and length is the number of code bytes
which are to be put out. Each call will put out one row of the
display. Unused bytes at the end of the stream (bytes left over in
the last word) should be zero. Any codes in excess of the maximum
number allowed on a line will be discarded.
It should be noted that this routine is nominally used for special
character sets, or for that matter, any character sets that are
software generated on dot-raster devices. In addition, however, it
can be used for photomosaic displays, area filling on maps, and
development of capability of producing audiovisual aids for
SCANLINE A raster-scanning display mode for which the
-------- user specifies a raster size, number of grey
levels needed, and direction of scan. The
subsequent data items specify only the location
and character of a change in the scanning
beam (or program).
The initializing procedure is SCANLINESET(DELTA, LEVELS, ORGMODE)
DELTA is an integer specifying the number of display points to be
included in each step. LEVELS denotes the number of intensity levels
to be used, and the sign of ORGMODE specifies whether the scan is to
be from the bottom up (plus) or top down (minus) on the display. For
both cases we will assume left-to-right. The absolute value of
ORGMODE gives the starting Y position.
All subsequent calls to this routine are of the form
The first such call denotes the origin in X and the initial
intensity. Subsequent calls denote the X value of the next point on
the scan where the intensity is to change, and that new intensity.
The program (or device) takes care of the stepping of the scan by
DELTA across the page, with the current intensity. Thus, the program
(device) only needs a data item for each change in the scan, not for
each position. When the next X is less than the previous X, or the X
position has been stepped to its limit, the Y position in the
incremented or decremented to continue the scan on the next line.
We see the device which accepts such a display as accepting a stream
of triplets of bytes, where the first two bytes (16 bits) specify the
X and the third (8 bits) specifies the level. The end of the stream
would be specified by three bytes of deletes (all ones). This
display mode is implemented in hardware on the display at the
Coordinated Science Laboratory at this University. It is the same
one which was used for the grey-scale work which has been reported by
SCANPOINT A point with intensity scanning mode in which
--------- the scan is handled automatically and only the
intensity of each point needs to be transmitted
to the program (device).
The initializing call for this procedure is
SCANPOINTSET(DELTA, LEVELS, ORGMODE)
The arguments are the same as for SCANLINE. The difference is in the
meaning of subsequent calls. The origin for the scan is at the left
end of the line corresponding to the absolute value of ORGMODE. The
stepping is done from left to right and at the end of each line, the
Y position is incremented or decremented by DELTA, according to the
sign of ORGMODE.
Subsequent calls to this procedure are of the form SCANPOINT(LEVEL)
where LEVEL denotes the intensity level at which the next point is to
be displayed. In this mode every point must have its intensity
specified by a separate call to the routine (byte to the device).
However, beyond the starting point no position information is
[ This RFC was put into machine readable form for entry ]
[ into the online RFC archives by Hamid Dastkar 09/99 ]