1 ©b. henderson-sellers semat 2010 semat position statement march 17, 2010 brian henderson-sellers...

1©B. Henderson-Sellers SEMAT 2010

SEMAT Position Statement

March 17, 2010

Brian Henderson-Sellers PhD, DSc, MASCE, FIMA, FACS, FIEAust, CPEng

Director, COTAR

Faculty of Engineering & Information Technology

University of Technology, Sydney

2©B. Henderson-Sellers SEMAT 2010

Overview

1st My “philosophy” for SE and SEMAT

2nd Some urgent issues in SE Method and Theory

3©B. Henderson-Sellers SEMAT 2010

Philosophy for SEMAT

Engineering applies science to real world problem solving

No theory of its own - relies on theories in other, more fundamental disciplines

4©B. Henderson-Sellers SEMAT 2010

Relevant established theories

• Category and set theory: mathematics• Estimation theory: statistics and signal

processing• Measurement theory: mathematics• Systems theory : ecology and

meteorology• Organizational theory: behavioural

psychology• Operational research: mathematics

5©B. Henderson-Sellers SEMAT 2010

Ignorance of existing theories is dangerous

Perils of ignoring theories

Tacoma Narrows Bridge collapse – a civil engineering lesson!

6©B. Henderson-Sellers SEMAT 2010

Task

Activity

Process

Life CycleProcess Group

Step

Activity

WorkDefinition

Task

Activity

Process

NoteList

(a) (b) (c)

+Child0..*

(d) WorkUnit

ProcessTechniqueTask

+StartTime+EndTime+Duration

+Parent0..1

+Context1

+Component0..*

SE has no agreed set of concepts

SE has no accepted terminology

7©B. Henderson-Sellers SEMAT 2010

In practice

• What is frequently called a process is just a set of ordered steps (a.k.a. procedure)

• Problem with many current ISO software engineering standards. (Arguably a process also includes resources, people etc.)

• Move away from Tayloresque factory processes to flexible processes that consider local situational context (risks, skills, culture etc.)

8©B. Henderson-Sellers SEMAT 2010

SE – still a proto-discipline

SE – today is an empirically-based proto-discipline has almost no data

Analogy can be drawn with the state of in business and management BEFORE their quality revolutions

9©B. Henderson-Sellers SEMAT 2010

SE Method and Theory

• My proposal

• Quantify and codify

Basics

Methods

Validation

Validation

Methods

BASICS

10©B. Henderson-Sellers SEMAT 2010

Start with basics

• Basics layer has a reliance on logic and mathematics - e.g. set theory, category theory, VDM

• Measurement framework is a vital element• Concepts (the “kernel language”) depicted

using formal mathematics or less formal visualizations based on this e.g. ontological structures, metamodels

11©B. Henderson-Sellers SEMAT 2010

Detour to metamodelling (one option)

endeavour

method

metamodelActivity

WorkUnit

Task Technique

* *

methodologies assessment quality tools

12©B. Henderson-Sellers SEMAT 2010

A metamodel

• is simply a model (of models)

• is often the core of CASE/CAME tools. Thus, implicitly accepted by users as being a “rule set”

• can provide an extensible framework across multiple abstraction levels

13©B. Henderson-Sellers SEMAT 2010

A simplified architecture

WorkProducts

Producers

WorkUnits

Stages

Languages& notations

create

aredocumented

using

produce&maintain

perform

providemacro organizationto the methodology

14©B. Henderson-Sellers SEMAT 2010

The ISO/IEC 24744 metamodel MethodologyElement

+Purpose+MinCapabilityLevel

WorkUnitKind

+Description

WorkProductKind

+Definition

ModelUnitKind

+Name

Template Resource

+Name

Language

+Name

Notation

+Expression

Constraint

+Description+MinCapabilityLevel

Outcome

EndeavourElement

+StartTime+EndTime+Duration

WorkUnit

+CreationTime+LastChangeTime+Status

WorkProduct

ModelUnit

+Description

GuidelineProducerKind

+Name

ProducerStage

StageKind

pow

erty

pes

15©B. Henderson-Sellers SEMAT 2010

Powertypes solve non-transitivity

project

method

metamodel

“MySystem”RequirementsSpecification

“MySystem”RequirementsSpecification

DocumentDocument

RequirementsSpecificationDocument

RequirementsSpecificationDocument

DocumentKind

DocumentKind

TitleVersion

TitleVersion

NameMustBeApprovedNameMustBeApproved

TitleVersion

TitleVersion

Req. Spec. DocumentMust be approved: yesReq. Spec. DocumentMust be approved: yes

“MySystem” Req. Spec.Version 1.5

“MySystem” Req. Spec.Version 1.5

16©B. Henderson-Sellers SEMAT 2010

More simply

Powertype pattern Resource (a metamodel class)

Clabject in method (a.k.a. process

model)

powertype “instantiation”

regular class instantiation

Metamodel domain

Method domain

Object in method

17©B. Henderson-Sellers SEMAT 2010

Part of a possible SEMAT metamodel

Universal

CompetencyActivityProducer Stage

StateGraphTask

Practice

Alpha

Pattern

WorkProduct

Ensure “Risk/Uncertainty” & “Quality” included

Mostly powertype patterns

18©B. Henderson-Sellers SEMAT 2010

Having established the basics

• How to create level 2 (Methods)

• Best candidate Situational Method Engineering

• SME uses method fragments, each defined in to be conformant to some Level 1 definition

• Ontological descriptions may be useful here

• Detour to SME - a good example

Project characteristics

Selection and Assemblyof Method Fragments

into Situational Method

Methodbase

Project characteristics

Selection and Assemblyof Method Fragments

into Situational Method

Methodbase

19©B. Henderson-Sellers SEMAT 2010

SME in a nutshell

Method fragmentsRepository

Methodology Instance

Step 2: Project Manager

Construction Guidelines

uses

Metamodel

conformant to

conformant to

Methodology M

Step 1: Method engineer conformant to

(e.g. ISO/IEC 24744)

20©B. Henderson-Sellers SEMAT 2010

Level 3 is also vital

• Engineering disciplines are observationally-validated

• Good quality data are vital to provide empirical evidence for “good practice”

• Accept validated concepts into “body of knowledge” (link to SWEBOK)

• Eschew “proof by assertion” – common today in SE

21©B. Henderson-Sellers SEMAT 2010

Outreach

• Agree and employ standards

• Propagate new vision

“Quantify and Codify”

Validation

Methods

BASICS