James Bo Begole

1
Flexible Collaboration Transparency

• James Bo Begole
• Department of Computer Science
• Virginia Tech

http//simon.cs.vt.edu/jamm
begolej_at_vt.edu
2
Collaborative Computing

• Shifting computer needs
• from Personal Computing to Inter-Personal
  Computing
• Computer support of cooperative work is emerging
• Some asynchronous (different time) support
• email, WWW, discussion groups
• Little synchronous (same time) support
• ICQ, NetMeeting

3
Software for Synchronous Collaboration

• Collaboration Transparency
• Application sharing
• Single-user legacy applications shared without
  developer awareness. The sharing is unknown, or
  transparent, to the application.
• Examples NetMeeting, SharedX, SharedApp, XTV
• Collaboration Awareness
• Applications designed for collaborative use
• Examples Editors (SASSE), whiteboards, chat,
  multi-player games (Diablo, Doom)

4
Collaboration Transparency
5
Collaboration-Aware Applications
6
Transparency versus Awareness
7
Problem of Concurrent Work

• User inputs can interleave and conflict
• Solution of conventional collaboration
  transparency take turns
• but that prevents concurrent work

(a) intended result of two users drawing curves
simultaneously
(b) unintended result of conflicting mouse
movement inputs
8
Sequential versus Concurrent
Sequential
User As site
User Bs site
Concurrent
9
What You See Is What I See

• Collaboration transparency
• strict WYSIWIS
• All participants have exact same view at same
  time.
• Collaboration awareness
• relaxed WYSIWIS
• Participants have independent views of shared
  data.
• Allows independent simultaneous work

10
Collaboration Transparency
11
Collaboration-Aware Applications
12
Group Awareness

• Information about participants
• identity
• location
• activity
• access

13
Centralized Architecture
User A Host
User B Host
User Input
User Input
Display
Display
Network Traffic

• One central copy of shared application
• Graphics output sent to each remote participant
• Remote inputs merged
• Higher network traffic

Conference Agent
Merged Input
Display Broadcaster
Conference Agent Host
Application
14
Support for Collaboration
15
Problem

• Conventional collaboration transparency has no
  extra development cost to share an application,
  but provides inflexible support of collaboration
• Why?
• Can it be improved?

16
Goals

• System Goals
• No extra development cost
• Lower network bandwidth requirement
• Usage Goals
• Concurrent work allowed in application
• Support multiple collaboration styles
• Familiar applications
• Relaxed WYSIWIS
• Detailed group awareness information

17
Flexible Collaboration Transparency
18
Approach

• Replicated network architecture
• Replace single-user interface objects with
  multi-user equivalents
• Optimistic distributed algorithm for concurrent
  text editing
• Automate turn taking where still required
• JAMM - Java Applets Made Multiuser

19
Replicated Architecture
User B Host
User A Host
Appli- cation
Appli- cation
User Input
User Input
Display
Display
Network Traffic

• Each person has a copy of the shared application
• Remote inputs are merged
• Lower network traffic
• Enables
• Independent views
• Concurrent work

Conference Agent
Event Broadcaster
Merged Input
Conference Agent Host
20
Late Joining

• Synchronize new-comer to current state
• Two approaches
• event replay
• image-copy
• JAMM uses image-copy
• Modified Java Object Serialization
• copies objects in current state

21
Proxied Externalities

• An externality depends on external state
• File, system clock, socket
• Cannot be replicated completely
• JAMM uses a semi-replicated approach
• Introduce replicated proxies
• Central server
• Communicate via Javas Remote Method Invocation

22
Proxied Externalities
Original Applet
DIS
FIS
23
JAMM
A single-user editor application containing a
scrollable panel ...
becomes a shared editor containing a multi-user
scrollable panel
24
Object Replacement
object reference
Multi-user equivalent objects are substituted for
single-user originals
25
Evaluation

• Flexible JAMM versus NetMeeting
• Two tasks
• Loosely-coupled collaboration
• Text Entry two authors simultaneously enter
  text
• Tightly-coupled collaboration
• Copy Edit editor leads author to make changes
  in a text

26
Performance Time

• Results
• Text Entry less time using JAMM (223.75 seconds)
  than NetMeeting (353.50) (plt.001)
• Copy Edit no difference, as expected (p 0.7905)

27
User Perceptions - Text Entry

• Q1 satisfaction with the software
• Q2 ability to work simultaneously
• Q3 ease of controlling the shared application

• Q4 ease of indicating text locations
• Q5 ease of simultaneous editing
• Q6 ability to have independent views
• Q7 ease of knowing partner's view

28
User Perceptions - Copy Edit

• Q1 satisfaction with the software
• Q2 ability to work simultaneously
• Q3 ease of controlling the shared application

• Q4 ease of indicating text locations
• Q5 ease of simultaneous editing
• Q6 ability to have independent views
• Q7 ease of knowing partner's view

29
Conclusions

• Flexible Collaboration Transparency
• No extra development cost
• Lower network usage
• Concurrent work
• Relaxed WYSIWIS
• Detailed group information
• Allows collaboration to flow between tightly- and
  loosely-coupled work
• Simple groupware toolkit

30
Questions?

• Why isnt JAMM twice as fast for Text Entry?
• Startup cost of coordinating who does what
• The completion time is bound by slowest typist