The Workspace Model - McGill School Of Computer Sciencejoerg/games/workspace.pdf · •Presenter has private view of presentation including current slide and associated notes •Presenter

The Workspace Model

T.C. Nicholas Graham, Greg Phillips,Chris Wolfe

Equis Group, Software Technology LaboratorySchool of ComputingQueen’s University

Geek Warning

How Do PeopleCollaborate?

• Spontaneously vs planned• With multiple threads of collaboration• Asynchronously vs synchronously• Formally vs informally• Collocated vs remotely• Using a variety of tools and devices

– physical vs virtual– PC, handheld, whiteboard, paper…

• Evolution in collaborative structure

Architecting CollaborativeSystems

• Conceptual architectures expressstructure of system– Components, connectors, actors, adaptors

• Implementation architectures exposeissues of distributed implementation– Concurrency control, networking, caching,

…

Architecting CollaborativeSystems - Problems

• Traceability – keeping conceptual,implementation architectures in sync

• Evolution – dealing with changes inarchitecture– Triggered by users– Optimizations– Faults

Workspace Model

Example Application

• Presenter has private view of presentationincluding current slide and associated notes

• Presenter has presentation flow controls• Audience members sit in an auditorium,

view presentation on a large screen, hearpresenter’s voice directly

Conceptual Architecture

Example Application

• Remote audience members viewpresentation on their PCs and listen to thepresenter via VoIP

• Remote/local audience members see sameslide

Conceptual ArchitecturePost Evolution

ImplementationArchitectures

• Expose implementation decisions– Allocation of components to nodes– Connector implementation

• Networking protocols, consistency maintenance

• Refinement relation maintains link toconceptual architecture

ImplementationArchitecture

Refinement

• Map conceptual architecture toimplementation architecture

• Recover conceptual architecture fromimplementation architecture– E.g., failure recovery

• Can be automated– Developer works only at conceptual level; runtime

system generates implementation architecture

• Expressed via graph grammatical productions

Example Refinement Rules

Evolution

• Systems change at runtime– New participant joins session– Participant changes location– Participant changes device– Network/node fails

• Evolution calculus is an algebra forexpressing such change– Implementable in toolkit– Give semantics of toolkit features

Example EvolutionCalculus Rules

Properties of WorkspaceModel

• Refinements are Local. Application ofrefinement rules is commutative.

• Refinements Terminate. Any non-trivialrefinement sequence will eventually lead toa ground architecture.

• Evolution and Refinement isCommutative.

Application: Restoral andRecovery

Application: WorkspaceSemantics of Extended C++

Status

• Workspace Model completely specified• Two toolkit implementations underway

– Automation of refinement relation• Evolutions at conceptual, implementation levels

– Plug-replaceability of implementation components• E.g., developer can choose consistency maintenance

algorithm to suit application

– Python– C++

• Natural C++ programming style• Support for existing code