Setting and maintaining a viable research agenda

John HoskingDepartment of Computer Science andCentre for Software InnovationUniversity of Aucklandjohn@cs.auckland.ac.nz

a = 2.b +c

NaOH + HCl -> NaCl + H2O

v = u + at

s = ut + ½ at2

• Visual languages• Design environments• Meta tools• Software engineering• Service oriented

architecture• Aspect oriented

design• Early aspects• Thin client interfaces• Mobile interfaces• Construction Eng• 3D Modelling

• Design patterns• Pattern languages• Process modelling• Data mapping• CSCW• Sketch and pen

based interfaces• Software visualisation• Knowledge mmt• Component based

systems• Cloud Physics• Medical Informatics

Notations + models• maths• chemistry• physics• electronics

Computer science• executable specns• simulation• languages & notations

Design support environments• engineering• software• design notations• expert systems

Platforms for specifying& building design environments• multiple views• multiple notations• consistency & constraints• code generation• notations for end users

Visual knowledge management• visual wikis• KM system spec and generation

Frameworks & meta tools forconstructing multi-viewmulti-notation environments

Design ToolsEngineering,Software

Ispel1991

MViews1993

JViews1997

JComposer/BuildByWire1998

Pounamu2003

Kea1991

Marama2005/6

ClassLanguage1988

VML/VPE1995

VBKE2007

ThinkBase2008

ThinkPedia2008

Visual knowledgemanagement &visual wikis

Frameworks forconstructing multi-viewmulti-notation environments

Meta tools for specifying &constructing multi-viewmulti-notation environments

Evolving Frameworks Pattern Language

Prolog Java Java + Web Services

Eclipse +Java

Applications developed using the frameworks & meta tools

Commercialisation/industry transfer

Ispel1991

MViews1993

JViews1997

JComposer1998

Pounamu2003

Marama2005/6

Ispel1991

MViews1993

JViews1997

JComposer1998

Pounamu2003

Marama2005/6

Which diagonal brace? Steel angle brace Let-in timber Pairs of flat steel strip Pairs of cut-in timber

Is the sheet material on: One Face Both sides

m. Plan Length of Brace m. Length Covered by Full-Height Sheet Material

m. Minimum Length of Brace

B.U.'s of Bracing Provided

Bracing element 3 of Bracing Line A

Help

1.7

external class DiagonalBracesInterface. determinant braceNo : integer. determinant braceLine : text. output minimumLength : metres. output bracingProvided : bu. output lengthError : text. procedure help. input diagonalBrace: [steelAngle, letinTimber,

pairsFlatSteel, pairsCutinTimber]. input sheetMaterial : [oneFace, bothFaces]. input planLength : metres. input sheetLength : metres. external procedure show. external procedure hide.end DiagonalBracesInterface.

class Bracing. generalisation DiagonalBracesInterface. parameter minimumLength, braceNo, braceLine. bracingProvided

:= planLength * sheetLength * buPerMetre. buPerMetre : integer

:= bu(diagonalBrace, sheetMaterial). lengthError

:= ‘The brace must be at least the minimum length.’if diagonal^planLength < minLength

| ‘‘. procedure help.

...end help.

...end Bracing.

Frameworks forconstructing multi-viewmulti-notation environments

Ispel1991

MViews1993

JViews1997

JComposer1998

Pounamu2003

Marama2005/6

MViews• Framework supporting construction of multiple view -

multiple representation design environments• 3-layer architecture

– Base– View– Display

• Changes 1st class• Used to implement

many example visualtools & environments

Display/ExternalLayers

ViewLayers

BaseLayer

class

class

generalisations

window

drawing_window text

text forms

features

class_icon

class_icon

gen_glueclass_text feature_text

text

drawing_window

window

external_class

External Interface(Data/Event interchange)

ExternalTool

...

feature

text forms

view rel. view rel.view rel. view rel. view rel.

Frameworks forconstructing multi-viewmulti-notation environments

Meta tools for specifying &constructing multi-viewmulti-notation environments

Ispel1991

MViews1993

JViews1997

JComposer1998

Pounamu2003

Marama2005/6

JViews:Reimplementation of MViews in JavaJComposer:

Visual tool for specifying the bottom two layers of the three layer JViews model – base (shared model) and view layers plus mappings between

BuildByWireVisual tool for constructing visual components of JViews toolsConstraint based

JViews & JComposer/BBW

Base CPRG

View Mappings

User Interface

+ Backend code gen

Orion Mapper Prototype

Meta tools for specifying &constructing multi-viewmulti-notation environments

Ispel1991

MViews1993

JViews1997

JComposer1998

Pounamu2003

Kea1989

Marama2005/6

Pounamu & Marama

a b

c

d

e

f

PounamuLightweight UILive changesThin & mobile clients

MaramaEclipse tool generatorExploits GEF, EMF, etcframeworks

• Platform versus project• Teams versus individuals• Weave in – spin out• Aim to be the best• Gain influence• Have fun

• Aimed at academiccareer, but applicablein industry research too

Some lessons learnt

Platform versus project• Build your research around a research platform

rather than a disconnected set of projects• Competitive advantage

– lower barrier for more complex projects

• Easier to get funding• Have a set of projects available that can

contribute to the platform• Be a magpie• Take a long term view & evolve the platform –

Evolving Frameworks Pattern Language

Time

3 Examples

White-box Framework Black-box Framework

Component Library

Hot Spots

Pluggable Objects

Fine-grained Objects

Visual Builder

Language Tools

Platform Migration

Application Integration

Reflective Framework

Original Evolving Frameworks Pattern

Language Design Patterns

Our Proposed Extensions to the

EFPL

Self-Extending Framework

Teams versus individuals• Don’t go it alone – look to collaborate

– Within your University– Outside

• You may need to adapt what you are doing to develop a shared platform

• Meet together• Co-supervise• Cross mentor

within the team– PhDs mentoring MScs– MScs mentoring project

students

Weave in – spin out• Weave new themes into your platform and

spin them out into new threads

Ispel1991

MViews1993

JViews1997

JComposer1998

Pounamu2003

Marama2005/6

IntegratedBuilding Design

ViewMappingLanguage

OrionSymphoniaMapper

VMLPlus

MaramaVMLPlus

FormBasedMapper

SoftArch

SoftArchMTE

ArgoMTE

MaramaMTE

PerformanceModelling

ArgoUML

FormCharts

LIDSSketch

SUMLOW

MaramaSketch

Aim to be the best• Don’t set out to be mediocre

• Strategies:– Learn from the best– Work with others who want to be the best– Pick a few things and focus on them rather

than spreading resources

Gain influence• You must get your message out

– Chase high ranked publication avenues– Look for PC memberships -> editorial boards– Get to know people – visit, collaborate

• Look for industrial impact (if appropriate)– Additional funding avenue– Good attraction for students– Typically need to adapt what

you are doing– Typically win-win

Have fun• Do things that you (and your colleagues &

students) enjoy doing

• If it’s too much like hard work do something else!

Acknowledgements• The many members of my research team &

external collaborators, but particularly:

John Grundy Santokh Singh

Rick Mugridge Hermann Stoeckle

Robert Amor Karen Liu

John Hamer Xiaosong Li

Richard Li Rainbow Cai

Nianping Zhu Jun Huh

Christian Hirsch Norhiyati Mohd Ali

Questions