Designing a multidimensional File Manager

• 1) Information organization
• 2) Metaphor design
• 3) Object data base
• 4) Self adjusting layouts
• 5) Layout schemes for a 3D type of file manager
• 6) Navigating in a 3D type of file manager

Note: The term ‘3D’ is replaced with ‘3D type of’ as there is confusion and misunderstandings on what 3D means when applied to UI design. 3D itself should not be the goal of the environment or design, but only one method to present data. A more concise definition will be proposed in a following paper.

Introduction

Designing a new type of user interface needs to cover many subject disciplines such as visualization, navigation, human perception, and spatial awareness. This outline addresses the major areas that need consideration for building a new type of non-immersive file manager with a goal of boosting the user's productivity. If a highly efficient new interface is properly designed, it will have a natural beauty that will speak for itself and not need to be explicitly implemented with extraneous visuals. The larger subject of metaphor design is briefly touched on as this alone warrants a separate paper.

A blended approach of visualization techniques is recommended for tackling the design of a new type of file manager, and this outline is somewhat generic as it is meant to be applied to any interpretation of a multidimensional interface design (e.g., 2.5D, 3D, 4D).

1) Information organization

Currently, there are two metaphors/models for file access and file management in current 2D desktops; one is the desktop interface, and the other being the file manager interface. As a result of the desktop interface not being integrated with the file manager interface, this leaves the user with two opposing methods for the management of folders and files. Here is a quick example.

Two methods for opening folders on current systems:

Spatial mode - Open a folder on the desktop, and a single window opens up with no hierarchical folder structure present, but the spatial position is the same as when last opened. Having the position of the window state retained gives the user a familiar relative position each time a folder is opened. (The spatial model is also called the object orientated model.) Pro: Simple and some use of spatial memory. Con: Extreme difficulty maneuvering deep folder hierarchies. ISSUE: WinXP implementation has problem losing positional information.

Navigation mode - Open a folder using the filer manager and the hierarchal tree lay out visualization is used to display the contents of the structure, with the contents usually displayed on the right side in 1-2 window panes depending on the platform. User navigates by clicking on little structural elements (plus and minus symbols) that opens and closes various directories. Pro: Okay mental model and simple concept to use up to 10,000 objects. Con: Constant clicking and scrolling required, limited ways for personal origination of presentation data. Data often presented with only one viewing method such as thumbnail or file details.

Without going into any further details on these two interface methods, the benefits of spatial location and the advantages of hierarchal tree layout need to be a combined to provide for a seamless user experience. This will lead to a consistent and universal method for accessing all types of objects (e.g., folders, files, icons, menus, favorites, media, and applications).

One other major area that is important for building a 3D type of file manager is what is loosely referred to as the “semantic desktop”. At its core is a small data base like scheme for tracking object relationships at a high level of detail. The term “semantic” is more or less a catch phrase that includes data base type of operations for automatic layout and theme related grouping of visual graphics. The data base is covered in major section 3.

Three broad organizational approaches to combine for a multidimensional user interface design:

1. Spatial and hierarchal layout models as mentioned earlier in this section. This also includes all 2D design elements and visualization effects such as tree layouts for navigation and cascading menu visuals. ISSUE: Bookmarks, shortcuts, and other custom grouping might need special schemes to be handled differently?

2. Semantic desktop concept – user centric driven with highly customizable graphics interface, self organization of local objects, semi/sub metaphor related, fixed and variable spatial layouts, theme / functional grouping capabilities (Windows Vista uses virtual folders.) User centric driven with an extreme range of personal layout preferences to suite a large range of users. ISSUE: Not sure where search function fits in as it needs to be integral, but this is much later in priority.

3. Object data base – use for tracking complex relational information between objects. This is the core of the semantic desktop above and is more of the abstraction I use to put the concept together. The object data base also calculates new positions, maintains the data base, and provides a pseudo API layer for high level distortions implemented by the global metaphor.

The term/phrase “semantic desktop” mentioned above in (2) is a general description to describe one of the organizational methods. This is different from the term ‘semantic web‘ which focuses around the organization of social networks. The commonality with these two terms is that both have to deal with a large amount of varying types of data, and this is fundamentally a organizational problem of the relationships between the objects and how they communicate changes when the environment changes. This touches on the issue of the metaphor. ISSUE: Better description of this loose bucket needed.

At its core, a new type of 3D user interface needs a common visualization method for uniformly accessing all data types such as files, folders, media, shortcuts, macros, and applications.

2) Metaphor design

To quickly touch on the subject of metaphor design, as with any new file manager, it will have to be an integral part of the global UI architecture. The global approach that advocated here is more about the design process and how to categorize problem areas. Also, from various research documents, the term ‘metaphor’ is also referred to as the model, distortion, focus + content, and the paradigm. These subtle differences usually depend on a document’s focus and the concept being promoted. ISSUE: Is a glossary needed?

The focus of this document is on designing a 3D type of file type of manager/browser and not on the global metaphor design itself.

To clarify two terms some useful terms, and to contrast a two tier way of thinking for designing a 3D type of UI, these two major definitions are asserted for future reference:

1. Global Object – Major effects like hyperbolic displays, major rearrangement of scale and size of windows, and other sudden focus plus content switching such as zooming, bending, or rotation of a large part of the display scene. I generally consider it a global distortion when one object affects another object.

2. Local Object – Local entities can be a file, a group of files, a file of any media type, a macro such as a shortcut, and perhaps a structural element. Any sub directory is also considered a separate local object, and any layout preferences.

Metaphor design is a lengthy subject in itself, but it can broadly be classify into two major metaphor types as reality based, and abstract type. All future references refer to the use of an abstract type of metaphor and not reality based metaphors (i.e., virtual worlds, planets, rooms, doors, hallways, or buildings).

3) Object data base

Caching and search go together well, and lately, these data base features have shown up in desktop designs (Apple with spotlight, Vista with Search Explorer), and integrating the search features throughout the system is a boost in productivity. However, the current methods for adding metadata to file/folders are for the user to better track, store, and organize personal files and groups, and are not for manipulation of the object layouts themselves. Higher fidelity tracking is needed on the position of objects.

What I’m proposing is a mini database that is internal to the design, and that works with the interface to track dozens of attributes per object. (Mostly local objects and it will probably operate in parallel with the current file system.) This will provide finely detailed information about the history and movement of the objects for future position calculations. I have a working design that uses basically a dynamic linked list where each object carries a table of all other related objects such as when one folder is inside of another folder. This provides a large amount of relational tracking information so objects can be automatically grouped and treated as a whole. ISSUE: Need diagram to clear up concept.

The algorithms that calculates positional information caused me some problems such as computing new locations when zooming in and out, but the data base architecture is straight forward to implement and is rudimentary for anyone versed in graphics design.

The data base approach offers great freedom for controlling objects in a multidimensional interface where quick calculations of future positions are necessary. In order to implement a complex and nested metaphor, it will need access to a wide range of virtual global objects. A data base approach is a core requirement of a new file manager design in order to provide high fidelity tracking.

The following object attributes need tracking for a flexible multidimensional type of environment.

• position in global, local, and virtual group
• visualizations available for the type of object
• position in virtual desktop (for 4D use)
• user preferences for viewing and grouping items
• history and pattern of viewing for self adapting
• screen space needed for maximum efficient viewing of object

The data base system will need to track dozens of object attributes, and it will need quick access to the objects for predicting new moves. Speed must be at a minimum of one frame per second (typically 1/30s). There are several ways to get started without over building this concept and this is the method that I currently use. Dividing up the object responsibilities could alter this approach, and it is only offered as a first start. This allows global tracking and analysis of when objects are moved or changed, and provides many inter operations between objects for self layout adjustment of data. ISSUE: Much more on this later. Also, how does LG track objects and operations?

4) Self adjusting layouts

The concept and definition of a local object is defined in section 2. Self adjusting layout of local objects is the most fundamental issue that needs to be implemented before any other part of a new 3D type of design can take place. It’s also the easiest part to implement as there are several solutions for displaying objects in nicely formatted rows and columns on the screen. Morphing transitions can also be added for good and useful visual transitions that don’t disturb spatial awareness, and to provide pleasurable use of graphics. This is a subjective subject, but from going over various usability reports, there is a level of “eye-candy” that is desirable as long as it does not slow down the user and can optionally be turned off or over rode. (This issue is covered later in section 6.)

Example of self adjusting objects – If you have 4 documents in a folder, and the containing window is maximized; I then want to see four quarter screen views of the actual documents. On the other hand, if I’m viewing 30 documents in a folder, then the display should present the items in a list box format. (Including use of heuristics such as last file used.) The idea is that the display will automatically adjust the layout based on many factors. This will provide the user with the optimum layout so as to minimize window thrashing.

The concept of groups and self adjusting layouts are under the heading of a structured design approach in that objects need to be constantly aware of their position and scale so they can self adjust their position and scale automatically to changing conditions. If the objects take into account both the global and local conditions, I believe that object occlusion/over-lap can be reduced to zero when navigating hierarchies/groups of data.

ISSUE: Needs better reasoning and examples. Also, what images can legally be included without copyright problem.

5) Layout schemes for a 3D type of file manager

There seems to be a few clever schemes for the lay out 3D data for viewing, but none of them lend themselves to a useful 3D type of user interface implementation which requires a tremendous amount of visual interaction. Many of these 3D/4D layouts use varying depth projections and basic visualization techniques for viewing standard 3D data at a few different vantage points with various clustering techniques, but interaction is limited as most of these designs target specific problems as typically seen in scientific applications.

Interesting multidimensional data visualization methods.

• hyperbolic distortion/display
• perspective wall, cone trees
• 4D graphing, virtual desktops, scatter plots
• partial hierarchal displays and automatic grouping methods
• spatial user interfaces  

ISSUE: Perhaps last bullet item needs its own break out?

Note on Hyperbolic displays – these schemes can handle a high amount of object nodes, and they give great visual feedback for testing methods of visual object manipulation in general. This type of display is a basic start for complex structural ideas, but not using hyperbolic geometry. I think there are good visuals for exploring new ideas.

However, most of the above approaches can’t meld with a user interface and here are several issues that need to considered for an efficient file manger design:

• prevent object occlusion when navigating so no overlapping of windows
• flexible navigation, but not too much freedom - solid land marks
• don’t use monolithic design which does not allow mixed mode visualizations schemes or sub dividing sections on the screen
• inner-communication between local objects for self adaptation
• complex nesting of objects for advanced layouts leveraging spatial capabilities

There are various tree types of layout visualizations that seem to offer the most flexibility for building structured layout arrangements. Tree type visualizations are easy to add to a design as they can be built in any dimensional direction like most graphic construction primitives. ISSUE: More experiments and user feedback on efficiency of maneuvering schemes. Samples needed.

6) Navigating in a 3D type of file manager

As mentioned earlier, the desktop interface and the file manager need to be combined so the interface is a seamless experience that incorporates the best of both models. This is the bases of a common visualization method for accessing all types of data.

I’ve been experimenting with 2.5D designs (zooming interfaces) where the windows stay perpendicular to the z axis, and I think this makes a good basis for a multidimensional user interface. Having a zooming routine that allows random drilling does not add value, but being able to control the layout of objects as the display is zooming in and out is where a balance is needed.

ISSUE: Tilt distortions/skewing applied to windows has problems when working with zooming type of interface, but a method for morphing the tilt back and forth to flat depending can be applied depending on where the zooming is at in the directory system. It needs to be turned off while actually navigating. Tilt effects works if you keep it at the top level such as on the desktop. This is a low priority compared to other parts of the design.

ISSUE: Need an example and better reasoning for the theory of using a 2.5D design. Make experiments and collect usability testing.

In order for efficient navigation to work, these categories also need addressing:

• simple structural elements need to be added for rigid way point navigation
• limiting the freedom of choices of layout and navigation
• user level control of user interface – levels
• inability to get lost by having a home plate to easily come back to
• future support of new HCI devices for better z axis control

One other note. The user should never have to wait for any visualization effect to be completed. Any mouse gestures on context sensitive objects that move or involve animated transitions should be linear in speed to the user’s movement so that it adapts to more advanced users and personal mouse gesture preferences. At all times, the transitions should have the ability to be over rode by the user so the task is instantly completed. This is more of a basic human perception issue and is related to human intuition and cognitive load.

(roberts – 6/25/06)