A 3D Window Manager

Papers CHI 2000 * I-6 APRIL 2000
The Task Gallery:
A 3D window manager
George Robertson, Maarten van Dantzich, Daniel Robbins, Mary Czerwinski,
Ken Hinckley, Kirsten Risden, David Thiel, and Vadim Gorokhovsky
M i c r o s o f t Research
One M i c r o s o f t Way
Redmond, WA 98052,USA
Tel: 1-425-703-1527
E-mail: ggr@microsoft.com
ABSTRACT
The Task Gallery is a window manager that uses interactive
3D graphics to provide direct support for task management
and document comparison, lacking from many systems
implementing the desktop metaphor. User tasks appear as
artwork hung on the walls of a virtual art gallery, with the
selected task on a stage. Multiple documents can be
selected and displayed side-by-side using 3D space to
provide uniform and intuitive scaling. The Task Gallery
hosts any Windows application, using a novel redirection
mechanism that routes input and output between the 3D
environment and unmodified 2D Windows applications.
User studies suggest that the Task Gallery helps with task
management, is enjoyable to use, and that the 3D metaphor
evokes spatial memory and cognition.
Keywords: Window managers, 3D user interfaces, spatial
cognition, spatial memory.
INTRODUCTION
Management of multiple user tasks is an activity that, if
made easier, could help enrich users' computing experience.
A task is a collection of documents and applications
organized around a particular user activity. For example, a
user may rapidly switch between working on finances,
writing a paper, and managing correspondence. Each of
these may involve many applications. Task management
has several components: creating, locating, and bringing
tasks into focus, and window management within a task.
Within a task, users need to manage placement and size of
windows and quickly shift focus of attention from one
window to another. Users also need to be able to bring
relevant information to bear on the task being performed. In
some cases, this requires bringing two or more windows
into a useful view simultaneously.
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies
are not made or distributed for profit oi' commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, to republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
CHI '2000 The Hague, Amsterdam
Copyright ACM 2000 1-58 ] 13-216-6/00/04...$5.00
Figure 1. The Task Gallery.
The Task Gallery is designed to meet the goals of task
management, while providing other features available in a
window manager. It is a 3D environment designed so users
can be productive using familiar, existing applications. Our
design premise is that 3D virtual environments can more
effectively engage spatial cognition and perception. Almost
all new personal computers are now delivered with 3D
graphics acceleration hardware. Although this innovation
has been driven by the computer game industry, it could
usher in a whole new class of productivity applications with
3D interfaces.
In the Task Gallery (Figure 1), the current task is displayed
on a stage at the end of a virtual art gallery. It contains
opened windows for that task. Other tasks are placed on the
walls, floor, and ceiling of the gallery. The user switches to
a new task by clicking on it, which moves it to the stage.
Viewing multiple windows simultaneously is done with a
button click, using automatic layout and movement in the
3D space to provide uniform and intuitive scaling.
Applications and frequently used documents are kept in a
Start Palette (Figure 6, described later) carried in the user's
virtual left hand. Studies suggest that users are enthusiastic
494 ~ k.~llili t~l=~X 2OO~2) CHI Letters volume 2 · issue 1
:HI 2000 ? I-6 APRIL 2000 Paper_s
about the Task Gallery, that it is easy to navigate the space,
and that it is easy to find tasks and switch between them.
The Task Gallery uses a novel redirection mechanism for
routing input events and graphics output between the 3D
environment and existing, unmodified Windows
applications.
PREVIOUS WORK
The window manager is the part of the computer user
interface that manages display and input device resources
[7]. It allows the user to bring up windows, menus, and
dialogue boxes associated with running applications,
manipulate windows, and minimize them. It takes mouse
and keyboard input and directs it to the appropriate
applications. The window manager also determines the
look and feel of much of the user interface.
window management systems have been a fundamental
part of computer user interfaces for the last 25 years. From
the mid-1970's to the mid-1980' s, there was much research
on window systems [2, 12, 13, 20, 21, 22]. By the mid-
1980's, Unix and MacOS window management had
converged on the desktop metaphor with overlapped
windows [18]. This metaphor has served the computer
industry well for 15 years, making it possible for many new
users to use computers effectively.
The desktop metaphor has changed little since it was
created. However, the way computers are used has changed
significantly. The growing range of applications and online
services have made computers applicable to many more
real-world activities. People often engage in a number of
tasks and need to switch between them frequently and
quickly [1, 9]. In the desktop metaphor, switching between
tasks can involve dozens of operations (iconifying,
opening, moving and resizing windows). Users often need
to see multiple documents simultaneously [10]. Again, this
can take many steps (opening windows, moving and
resizing them, and scrolling). The desktop metaphor has
inadequate support for task switching, leading to wasted
effort and frustration on the part of the user [4, 9, 10, 16,
17]. In the Task Gallery, switching between tasks and
viewing multiple windows simultaneously are simple
actions. In addition, the Task Gallery provides a strong
spatial framework for encoding location information and
front to back relationships, thereby engaging the user's

spatial memory to help retrieve tasks and services.
Rooms [9] was created to deal with problems that early PC
users had in managing their tasks [1]. Users switch between
tasks frequently and there is strong locality of window
reference based on task: particular windows are associated
with particular tasks. This can be exploited by creating a
visible representation of a task, and allowing the user to
easily switch between tasks (which combines opening,
sizing, and placement of multiple windows into one act).
The Task Gallery takes advantage of the user's spatial
memory for task management, while Rooms lays out tasks
in a linear alphabetic order. The Task Gallery currently
lacks one feature of Rooms, which is the ability to share a
window so that it appears in multiple tasks; however, we
plan to implement this in the future.
The Andrew window system [12] explored a space-filling
tiled window layout, where windows are resized
automatically (when one window grows others shrink by
cropping to keep the space filled). Users found it confusing
and the approach was abandoned.
Elastic Windows [ 10] also uses a space-filling tiled layout,
with tasks replaced by hierarchical user roles. The lowest
level role is similar to the Rooms notion of task. The Task
Gallery returns to tasks as the basic unit, and uses spatial
layout of tasks for task management, instead of a role
hierarchy. Elastic Windows addresses the problem of
simultaneous display of multiple windows by allowing the
user to create a container into which multiple windows can
be dragged. The Task Gallery has similar functional
advantages, but no special container is needed and only a
single button click is required to select each window. The
Task Gallery also maintains spatial continuity whereas
Elastic Windows can do significant window repositioning.
3D Rooms [16] was built as an information workspace that
used 3D virtual environments to extend the ideas of Rooms.
This was not actually a window manager; abstract
information visualizations replaced windows. The basic
motivation was to engage human spatial cognition and
perception in order to make task management easier. The
Task Gallery shares that motivation, but manages windows
associated with existing applications in order to bring the
advantages of human spatial cognition and perception to
our current set of computer applications.
Web Forager [4] and Data Mountain [17] are virtual
environments designed for managing documents. They
each use a 3D virtual environment to more fully engage
human spatial cognition and memory. Studies of the Data
Mountain [5, 17] demonstrate that placing documents in a
3D space helps the user remember where the documents are
during later retrievals. The Task Gallery also seeks to use
spatial memory to help the user remember where tasks are
placed in the gallery.
TASK GALLERY DESIGN
The choice of a navigable spatial metaphor was partly
motivated by a desire to leverage human spatial memory
[17]. An art gallery was chosen because of its familiarity.
To increase ease of retrieval, the Task Gallery includes the
images of documents and tasks in the space in addition to
their spatial location and title cues. Classical mnemonic
research has documented that mental cues in the form of
visual images are an excellent way to enhance memory for
items [14]. Our previous studies have shown the strong
influence of snapshot/thumbnail cues to aid spatial memory
during the storage and retrieval of web pages [5].
~k.~ll~l 495
Papers CHI 2000 ? 1-6 APRIL 2000
Look Up
Look to Left ~ ~ Look to Risht
~t" LG~ DF°r~Wdlr
G° ~a~km0~~Z~S~h~ me View
Overview
Figure 2. On-screen 3D navigation controls appear
in the lower left corner of the screen.
Animation is used to reinforce the spatial metaphor. For
example, when a user switches to a task by clicking on it, a
one-second animation starts. The current task is closed by
creating a snapshot which is moved back to the task's frame
in the gallery. The newly selected task is then moved from
its frame to the stage. When it arrives at the stage, it is
transformed from artwork into live windows. A "ghosted"
view of the task remains in the gallery, to mark the spot that
it came from.
The initial and primary working view is a close-up of the
stage (Figure 4), showing the current task and its live
windows. To view other tasks, the user backs up to see
more of the gallery, as in Figure 1. The gallery is composed
of a sequence of rooms, with only one closed end; more
rooms are revealed without limit as the user moves back.
This provides a simple way of managing the user's
attention. As the user backs away, attention is widened.
Moving to the stage focuses attention on the current task.
The user can move tasks wherever desired with a dragging
movement. Tasks are constrained to remain on walls, floor,
or ceiling, but can be moved between these surfaces in a
way inspired by Object Associations [3]. The transition
from wall to floor, for example, causes the task to shift to
the appropriate orientation on the floor. Task frames are
tilted outward so that they are more legible from a distance.
Task frames on walls are mounted on a stand to make the
metaphor more obvious and to ground them visually in
depth. Segmentation of the gallery into separate rooms,
grouping of task windows into mounted pieces of artwork,
and using distinctive backgrounds all provide landmark and
spatial cues that act as memory aids.
Users (especially non-garners) tend to get lost in many 3D
systems that require them to navigate. We avoid this
problem by keeping the space simple (a linear hallway), by
choosing a metaphor appropriate for the context (viewing
Move Bring Ordered Loose Add to Maximize Close
Forward, Stack Stack Selection
Send back
Figure 3. Window manipulation controls appear over
a window banner when the user points to it.
art in a gallery), and by constraining the navigation. Thus,
we provide a few simple controls rather than a general
egocentric navigation mechanism. Figure 2 shows these onscreen
controls, which allow the user to "jump" backward,
forward, home (primary view), and to a bird's eye view
showing all the tasks in the Task Gallery. Each jump
control starts a one-second camera animation from the
current position to the desired target. Our studies showed
that users did not become disoriented in the 3D space when
using these controls, and that they could easily find their
desired tasks.
New tasks can be created by picking the "new task" item on
a menu or on the Start Palette (described later). A
background image is chosen by the system to distinguish
the new task from existing tasks. The user's desired location
of the new task is not yet known, so it is placed on the floor
in front of the stage. Other tasks on the floor are moved
back away from the stage to make room for the new task.
This is done with a three-step animation: move the camera
back to make the action visible, move the tasks on the floor
back and place the new task on the floor, and finally do a
task switch as described earlier. The three-step animation
was implemented as a result of user testing, and greatly
improved the usability of task creation. It is assumed that
the user will move the task to a more appropriate location
in the gallery later.
window management
The current task on the stage is composed of several
components, including a loose stack, an ordered stack, and
a selected windows set. The loose stack is used for
overlapped windows in much the same way as the current
desktop metaphor. These windows are mounted on stands
to visually ground them to the stage. Clicking on one of
these windows will bring it forward to a selected window
position, replacing the current selected window. The
window manipulation controls shown in Figure 3 are used
for moving windows around and placing them on various
stacks. These controls appear over the window banner
when the user points to the banner. Windows in the loose
stack can be directly moved anywhere on the stage, using a
technique similar to Point of Interest object movement [11];
mouse movement controls movement in the plane
perpendicular to the line of sight, and the shift and control
keys control movement toward or away from the user.
496 .~k.~i~
~I=IZ 2 0 0 0
CHI 2000 ? 1-6 APR|L 2000 Papers
Figure 4. Ordered stack, one selected window.
The ordered stack appears to the left of the stage, as shown
in Figure 4. Users choose to place windows in the ordered
stack to keep currently unused windows organized (e.g.,
open email messages). If one of the windows on the
podium is moved, the stack is tidied to have a fixed
distance between each window. Clicking on a page in the
ordered stack moves it to the selected window region.
When windows are selected, the system moves them closer
to the user for greater legibility. multiple windows can be
selected using the "Add to Selection" icon in Figure 3. Each
time a window is added, an automatic layout moves the
windows so they are all visible side by side (Figure 5).
Unlike tiled window managers that crop windows and may
force users to scroll, this operation does not affect what is
visible in selected windows. Thus we use distance in 3D to
provide uniform scaling in an intuitive way.
Toolspaces
The existing Windows desktop metaphor uses menus
(especially the Start Menu) and toolbars to give the user
access to commonly used tools and documents. To better fit
the metaphor of moving through a hallway, we designed the
Task Gallery so that the user carries tools and documents
associated with the virtual body, using an adaptation of
Glances and Toolspaces [15]. Glances are a lightweight,
ephemeral way of looking around in a virtual environment
without moving the virtual body. Toolspaces are placed
around the user, and hold various tools or objects, keeping
them associated with the virtual body as it moves through
the virtual environment.
The Task Gallery has toolspaces left of, right of, above, and
below the user. Hands and feet are shown to make the scale
of the objects in the toolspaces more obvious and to suggest
that these tools stay with the user as the user navigates
through the environment. In the Task Gallery, glances are
initiated with the controls shown in Figure 2. Glances
remain in effect until the user selects something in the
toolspace or glances elsewhere.
Figure 5. Multiple selected windows.
The Left toolspace contains the "Start Palette", which is a
Data Mountain [17] with the appearance of an artist's
palette (Figure 6). The Data Mountain was originally a
tilted plane in 3D holding favorite web pages. The objects
on the Start Palette are icons and snapshots for applications,
favorite documents, or web pages. The behavior of the Start
Palette is similar to a Data Mountain, including object
movement and occlusion avoidance. The only difference is
that selecting an object from the Start Palette causes an
application to launch with its window(s) in the current task.
When an application is launched, the glance is terminated.
Our user testing demonstrated that participants learned to
add applications and documents to their tasks easily using
the Start Palette. Earlier studies of the Data Mountain [17]
suggest that users should be able to find icons on the Start
Palette much faster than in the traditional Start Menu.
TASK GALLERY USER TESTING
During the design and implementation of the Task Gallery,
we gathered empirical evidence to support our design
decisions. Our first three studies examined task
management before and after various usability issues were
resolved. The third study took place several months after
the first two, and included evaluation of features added in
response to the first two studies (e.g., icon identification).
In addition, we were interested in how spatial cognition
pertains to 3D environments like the Task Gallery, and
whether or not aspects of real world spatial location
memory transfer to electronic environments.
We were interested specifically in how well users could
create and modify tasks and arrange the overall task space.
In addition, detailed information about organizing and
retrieval strategies was collected, to support those strategies
in future designs. We wanted to know whether organizing
strategies were based on frequency, size, type of content or
time. While the art gallery metaphor suggests use of the
walls over the floor and ceiling, previous research suggests
that certain bodily axes are considered primary in the real
world [8, 19]. We wanted to know if participants'
organizing strategies and subsequent retrieval performance
~k.~l~ 497
Papers CHI 2000 ? 1-6 APRIL 2000
and representation of the space related to properties of the
metaphor or to up-down, front-back and left-right axes
relative to the user's orientation in the space.
We also wanted to know how subjectively satisfying
participants considered the 3D user interface for task
management. A benchmark comparison between the Task
Gallery and Windows is planned after further design
iteration.
Experiments I and 2mPrototype System
Method
Participants
Eleven participants (5 female) between the ages of 16-65
participated in two iterations of the same study. All were
intermediate to expert Windows users. Five participants
evaluated the first iteration prototype, and six participants
evaluated the second iteration prototype.
Materials
Materials included two prototype versions of the Task
Gallery, which used "snapshots" of documents instead of
live applications. The prototype environments were fully
interactive except that the applications were not live. Eight
tasks and their contents were created prior to the study,
based on common computer tasks collected during actual
Windows' user home visits. Tasks typically contained 2 to
5 documents (like Word documents, Excel spreadsheets,
web pages, and email). Images of documents comprising
these tasks were saved onto the Start Palette in a default
arrangement used for each subject. During the study,
participants used the Start Palette for items to add to tasks
and to create new tasks. The Start Palette had 33 items.
The study was run on a 300Mhz Gateway Pentium
computer with a (1024x768 resolution) 15" NEC Mutisync
LCD flat panel monitor. P.a~. icipants interacted with the
software using a standard Microsoft serial mouse. No audio
was included in this prototype:.-
Procedure
Participants carried out 6 tutorial trials, and 20
experimental trials. In the tutorial trials, users were
introduced to the concepts of navigating, selecting, and
arranging documents and tasks in the environment. Once it
was determined that the participants could perform all of
the tutorial trials easily, the experimental trials were begun.
During the experimental trials, users created tasks,
organized the tasks in a way that was meaningful to them,
retrieved eight tasks and their specific content items, and
finally carried out various Windows operations. After the
first experimental trial, we asked users to draw what the
hallway looked like to them, and what location and
orientation they had within the hallway. At the end of the
session, users drew their information layout in the hallway
in as detailed a manner as they were able. In addition, they
filled out a user satisfaction questionnaire.
Figure 6. Start Palette - A Data Mountain held in
the user's left-hand toolspace.
Between the first and second study, several changes were
made to the prototype in response to observed user
problems. We changed the manner in which tasks were
created, named, and labeled when selected.
Results
Trial Times
Trial times for each subject were averaged across trials for
each trial type in the experimental phase of the study.
Overall trial times improved after changes were made to the
user interface by about 7 seconds (range 25.4 to 9 seconds),
on average. None of the performance improvements
reached statistical significance due to the small number of
participants and the large individual differences observed.
Organizational Strategies
Participants placed significantly more tasks on the left and
right walls of the gallery than the ceiling or floor. On
average 4.18 of the tasks were placed on the walls while
2.18 were placed on the ceiling and floor (_t (10) = 2.54, p window manager
so that bitmap sharing is efficient. For example, Feiner [6]
modified an X server to put 2D windows into a 3D
augmented reality.
The Task Gallery runs on current high-end PCs with a
modified version of Windows 2000 and a standard 3D
graphics accelerator (NVidia TNT2). All Task Gallery code
was implemented in C++, using Win32 and Direct3D APIs.
Task Persistence
One key problem which Rooms [9] faced was the capture
of information necessary to persist the state of tasks with
running applications, so that on restart all of those
applications are re-launched and the user sees exactly the
same layout last seen in each task. The Task Gallery faces
the same problem. The best we can currently do is to record
the information used to launch the application.
Unfortunately, that is far from ideal. Applications allow the
user to change what files are open, and some even provide a
form of window management within the application.
Without some standard way of getting the state of open
files and sub-windows within an application, it is extremely
difficult to solve the general problem. Some Windows
applications allow inspection of their open documents
through COM interfaces. We are exploring what can be
done by tracking file opens and closes and window
creation, but this approach is difficult without modifying
existing applications.
DISCUSSION AND FUTURE WORK
The Task Gallery is an exploration of the use of 3D virtual
environments for window and task management. It is
motivated by the desire to leverage human spatial cognition
and perception and to take advantage of the coming
ubiquity of 3D graphics hardware for more than computer
games. Early user tests suggest that the Task Gallery does
help with task management and is enjoyable to use. But we
have only scratched the surface.
In our usability studies we observed users exhibiting many
of the same principles of spatial cognition as are exhibited
in the real world. Users had a strong sense of front to back
ordering of their tasks, rarely confusing that ordering in
memory. We will continue to explore metaphors leveraging
users' real world knowledge in our future 3D environments.
There are a number of things that we plan to do as we
continue to evolve the Task Gallery. Better landmarks
could make a significant difference in helping users
remember on which wall they placed tasks. The Data
Mountain occlusion avoidance algorithm can be used to
help avoid occlusion problems while moving task frames.
As discussed earlier, the task persistence mechanism may
benefit from application-level changes, although we hope to
avoid those. These changes will make it possible to
effectively use the Task Gallery as a replacement for the
current desktop on a day-to-day basis. Once these necessary
changes are made, we intend to do a benchmark study
comparing the windows desktop shell with the Task
Gallery. Beyond that, we plan to explore integration of
novel uses of 3D visualizations living side-by-side with
existing Windows applications.
Our goal was to design a 3D window manager that solves
two problems with the current desktop metaphor: task
management and comparison of multiple windows. The
Task Gallery is a first-generation system that addresses
these issues, and is built on a general-purpose application
redirection technology which will allow us to explore
alternative user interfaces for application environments.
5OO ~
C;I=IZ ;:~'C)OO
CHI 2000 ? 1-6 APRIL 2000 Papers
ACKNOWLEDGMENTS
We thank Matt Conway, Susan Dumais, George Furnas,
and Dennis Proffitt for valuable assistance during the
design of the Task Gallery. The Windows 2000 USER and
GDI teams (primarily one of the authors, VG) were
instrumental in architecting appropriate low-level support
for this work; we owe great thanks in particular to Corneliu
Lupu and Andrew Goossen.
REFERENCES
1. Bannon, L., Cypher, A., Greenspan, S., Monty, M. L.,
Evaluation and analysis of users' activity organization,
CHI'83, ACM, New York, NY, (1983), pp. 54-57.
2. Bly, S., Rosenberg, J., A comparison of tiled and
overlapping windows, CHI '86, ACM, New York, NY,
(1986), pp. 101-106.
3. Bukowski, R., and Sequin, C., Object associations: a
simple and practical approach to virtual 3D
manipulation, in Proceedings of Symposium on
Interactive 3D Graphics, 1995, pp. 131-138.
4. Card, S., Robertson, G., York, W., The WebBook and
the Web Forager: An information workspace for the
World-Wide Web, CHI96, ACM, New York, NY,
(1996), pp. 111-117.
5. Czerwinski, M., van Dantzich, M., Robertson, G., and
Hoffman, H., The contribution of thumbnail image,
mouse-over text and spatial location memory to web
page retrieval in 3D, Proceedings oflnteract '99.
6. Feiner, S., Maclntyre, B., Haupt, M., and Solomon, E.,
Windows on the world: 2D windows for 3D
augmented reality, UIST '93, ACM, November 1993,
pp. 145-155.
7. Foley, J., van Dam, A., Feiner, S., and Hughes, J.F.,
Computer graphics, principles and practice. Addison
Wesley, 1996.
8. Franklin, N., & Tversky, B. Searching imagined
environments. Journal of Experimental Psychology:
General, 199, (1990), pp. 63-76.
9. Henderson, A., Card, S. K., Rooms: The use of
multiple virtual workspaces to reduce space contention
in a window-based graphical user interface, ACM
Transactions on Graph&s 5, 3, (1986), pp. 211-243.
10. Kandegon, E., and Shneiderman, B., Elastic Windows:
Evaluation of Multi-Window Operations, CH1'97,
ACM, New York, NY, (1997), pp. 250-257.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
Mackinlay, J., Card, S., and Robertson, G., Rapid
controlled movement through a virtual 3D workspace,
SIGGRAPH '90, ACM, 1990, pp. 171-176.
Morris, J., Satyanarayanan, M., Conner, M., Howard,
J., Rosenthal, D., and Smith, F., Andrew: a distributed
personal computing environment, CACM, 29(3),
March 1986, 184-201.
Myers, B., Window interfaces: A taxonomy of window
manager user interfaces, IEEE Computer Graphics and
Applications 8, 5 (September 1988), pp. 65-84.
Patten, B.M. (1990). The history of memory arts.
Neurology, 40, 346-352.
Pierce, J., Conway, M., van Dantzich, M., and
Robertson, G., Toolspaces and glances: storing,
accessing, and retrieving objects in 3D desktop
applications, in Proceedings of Symposium on
Interactive 3D Graphics, April 1999, pp. 163-168.
Robertson, G., Card, S., and Mackinlay, J., Information
visualization using 3D interactive animation, CACM,
36, 4, (1993), pp. 57-71.
Robertson, G., Czerwinski, M., Larson, K., Robbins,
D., Thiel, D. & van Dantzich, M.. Data Mountain:
Using Spatial Memory for Document Management,
UIST '98, ACM, November 1998, pp. 153-162.
Scheifler, R.W., and Gettys, J., The X window system,
ACM TOG, 5(2), April 1986, 79-109.
Siegel, A., & White, S. The development of spatial
representations of large-scale environments. In H.
Reese (Ed.), Advances in child development and
behavior, 10, (1975), pp. 9-55. NY: Academic Press.
Smith, D.C., Irby, C., Kimball, R., and Harslem, E.
The Star user interface: An overview. In Proceedings
of the National Computer Conference, 1982.
Teitelman, W., Ten years of window systems - a
retrospective view, in Hopgood, F.R.A., et al., eds.,
Methodology of window management, Springer-
Verlag, New York, 1986, pp. 35-46.
Tesler, L., The Smalltalk environment, Byte, 6, 8,
August 1981, pp. 90-147.
~,~.~
~k~l~ 501