ASWEC 2009 Wrapup

Earlier last year I attended the Australian Software Engineering Conference (ASWEC) on the Gold Coast. I recently had to dig up my notes when creating a summary of my year in Agile (2009), and, since the conference was before I started blogging in anger, I decided to post up my notes now.

My impressions of the conference was that whilst its theme was “Agile, the New Mainstream”, many of the core attendees (mainly from the academic and research environments) were relatively unaware of agile in practice. My core criticism was that unlike other conferences I have presented at, this conference requires the speakers to still pay to attend. Luckily my employer was a sponsor and negotiated a free pass, but for this reason I did not submit a proposal for the ASWEC 2010 conference that is being held in Auckland, New Zealand.

The following are my notes from the sessions that I attended.

Keynote: Specialization and Economies Of Scale

Mary Poppendieck delivered the day 1 keynote:

deal with complexity? – divide and conquer – along what lines do we divide?
“The Pin Factory” (Adam Smith, 1776) – divide by specialisation
1930, Ford – the ultimate pin factory – variety is the enemy – a car for every purpose
cost of changeover – batch and queue – $ cost to change machines over, Little’s Law (queuing theory)
cost of delay – cash flow, obsolecence
cost of mistakes
cost of motion
cost complexity – scheduling, learning
use a work cell instead of economies of scale – relied on cross functional team
1968, Edsker Dijkstra defined structured programming and proclaimed need to eliminate bugs to start with, the programming will be cheaper
1972, Harlan Mills defined top down programming – should not have difficulty at integration
divide by hierarchy / organisation
1972, Dave Parnas defined criteria for decomposition – modular programming, easier to change
1974, Vint Cerf / Robert Kahn invented TCP/IP – each network stands alone, no information retained by routers, no global control, communications on best effort basis
division by matching the domain

Amazon – all about scale

transitioned to services, each one
owned by two pizza team – must be able to feed the team with two pizzas
eliminated communication between teams, seven people cradle to grave, cross functional, stay a minimum of two years
now sell as cloud computing
like TCP/IP, don’t need to reach agreement – eventual compatibility, don’t worry about different databases, accept and deal with failure
software structure meets organisation structure
teams need independence to be creative
Conway’s Law

1976, Barry Boehm – SDLC – specialisation (like Ford)
IBM – convert 35,000 engineers world-wide to agile
WebSphere – 10 month deadline, customer feedback every iteration, early access program, customer feedback on discussion forum – 50% of marketing information wrong so developers listened to customer – phenomenal sales day one of release, support calls down by an order of magnitude
case study – do agile right and find defects the minute they happen
xUnit testing – one of the best software engineering advances in the last decade
batch and queue so ingrained, we fail to see queues of work that will take years to clear
utilisation paradox – 100% utilisation of highway is a parking lot
Empire State Building – September 1929 to May 1931 – on time and 18% under budget, demolition to finish – biggest issue was supply of materials, the only thing that was scheduled (Carol Willis, “Building The Empire State” builders notebook)
4 pacemakers to build the building
schedule the constraint – use a pacemaker schedule for the constraint

Towards Specification Based Testing for Semantic Web Services

Gill Dobbie delivered this presentation:

most people test against the web service specification – Coyote, WSDLTest
need to test web service against user goals
WSMO – Web Service Modelling Ontology

Intellectual Property In The Software Industry

I always enjoy presentations on the legal side of IT, and Alistair Smith delivered this informative presentation:

4,000 patents last year to IBM, USA has the most number of patents
copyright protects the expression of an idea, not the idea itself
patents protect products, methods or processes – jurisdictional (so need to apply multiple times)
confidential know-how – not known to the public (eg. the recipe for Coca Cola)- work must be secret, can be leaked
trademarks – distinguish goods / services of a trader, Coca Cola 80% is recognition of product – must be distinctive, reproducible, people trading for smells (tyres that smell like roses), can be renewed indefinitely
strong protection – need range of IP types
do an R&D review eg. Kambrook and power boards
patent pending – filed but not yet granted
open source and patents do not mix – but use other IP
reward inventors
formalise IP ownership – make explicit in employment contracts
review existing IP – company that had patent for JPEG compression missed 10 years of royalties
maintain records of R&D – evidence of inventorship, ownership and prove date of creation for copyright purposes – maintain previous versions of software in unchangeable form, keep notes in bound notebook, dated and signed
search IP databases – look at what competitors are doing
use trademarks that are registerable – not descriptive of product, should be distinctive
don’t just rely on copyright – does not protect against independent creation
don’t rely on confidentiality agreements alone – provides very little protection and hard to prove and hard to recover from
don’t rely on open source software being unencumbered by patents eg. Microsoft has 200 Linux patents, IBM “bluewash” on its new takeovers and check code line by line (takes six months)
no 10% rule on copying software – infringe is a substantial part – 2 JS/CSS files in a content management system
don’t forget to mark products – impact on damages

Invited Presentation: Bringing Agile To Life At The Enterprise Level

Jeff Smith, the CIO of Suncorp, delivered this keynote presentation:

feedback is essential
Wikipedia – 6 employess, 200,000 hours of contribution from community
Grateful Dead – high grossing band, found people like listening to music together, “collaboration”
leaders need to encourage simplicity
provide tools to enable collaboration
leaders use passion not threats to get things done
innovation / entrepreneurship is for everyone – need to channel the passion
Pixar – core beliefs, simplicity, intellectual capability
managers need to help teams get out of failure
trusted environment – safe to tell the truth, create a learning environment
creativity for everyone not just the artists, focus on people not the ideas
disseminate the authority to the teams
similarities between Pixar and agile
get people interested in learning
why? why now? why do it this way?
we are good at reacting, not so good at responding, initiate change is hard
generosity is hard – right for community before yourself
managers have employees, leaders have followers
culture top down, innovation bottom up
fear of failure – actually fear of blame / feedback, very few people fired due to a failed project
worse than “no” is a “not yet”

Can change behaviour but hard to change beliefs:

don’t believe what you tell them
don;t believe what you show them
believe what friends tell them
people always believe what they tell themselves

recognise leadership is scarce – is uncomfortable
few things happen by accident
don’t settle, have faith in what we do
inspire, connect, leverage…

Agile In Government: Successful On-Time Delivery of Software

Adrian Royce delivered this presentation (and the paper won the award for best industry paper):

agile buzzword – everybody wants to be flexible
chose Scrum – collaboration and for the whole team, encompasses agile principles
backlog, only tasks of highest immediate business value were actioned
pilot project successful (Housing)
twelve successful projects delivered using Scrum (6 = pilot, 3 < pilot, 3 > pilot)
largest Scrum project undertaken ($3 million, 4 months ahead of schedule)
not everyone ready for Scrum, “just common sense”
sense that agile projects increase risk
agile = no documentation – government audits require documentation
obstacle – organisation and culture
benefits – hard to quantify, able to respond quickly
increase in morale, productivity, adaptability and business engagement, low staff turnover
Scrum Master is essential, focus on team, remove impediments, sprint zero is a necessity
do not impose, keep productive teams together, human interaction over technology
measure everything, focus on quality (client was not going to be the first to find a bug), leadership essential

Implementing an Enterprise System at Suncorp Using Agile Development

James Couzens delivered this presentation:

what worked well – business engagement, wide stakeholder group
not working well – coordination with integration team, managing backlog, overly focussed on showcase
definition of done not appearance of done
textbook agile does not scale without adjustments
business need to be onboard

Keynote: Agile Thinking – From Concept to Reality

Graeme Wood, the founder of Wotif, delivered this keynote:

started using agile without knowing they were using it
design constrained by six UI screens and a rough database design
chaos is fun, makes people think quickly
market for hotel rooms not working efficiently (travel agents) – only 60% – 70% occupancy
no business plan, no design, originally written by two people sitting side by side, allowed for quick change
launched in 2000 with 60 hotels, slim margin
no advertising budget, word of mouth, wined and dined journalists, wrote stories, etc
simplistic UI – has hardly changed
supplier can change cost hour by hour – not locked in
fewest number of clicks, no login / registration
keen not to have hospitality people – that was model we were trying to change
small teams, 9-10 max – largest system made them focus on the architecture – if don’t have that becomes “wild west territory”
flat organisational structure, keep small teams focussed on what they are there for
business dragged back from solving all of the “distressed inventory” opportunities
need to get agile thinking into the business planning area why don’t more IT people go towards management? – IT such a big part
start your own conversation, send a demo to influence, start a future systems blog
communication with potential business partners or customers
social networking has opened up communication outside normal business
all this fails, resign and start your own business
majors came aboad reluctantly, no paid positions, alphabetical is star rating, started renaming properties and getting regraded – leveled the playing field
suppliers to hospitality industry have been slow to change
started with Visual Basic, now all open source
Wotnews – public site plus corporate solution to look for customers, staff, etc… Initially around financial services, feeds from the ASX, verging on Media Monitors, greater depth than Google News, can get Wotnews to get competitors newsletters
re-architected about two years in, cancelled big bag because it didn’t feel right,now doing incrementally

Static Bug Checking Tools and Why You Should Care

Nathan Keynes delivered this presentation:

date back to Lint (1978)
FindBugs
most people just run once
most bug checkers report false positives and false negatives
commercial bug checkers expensive, comparisons not available and price on size of codebase
upfront effort, legacy codebase may not be parseable
Parfait for C++ / C – Sun for Solaris
BegBunch – benchmarking bug checking tools – accuracy / performance

Simulating Software Evolution with Varying Numbers of Developers and Validation Using OSS

Steve Counsell delivered this presentation:

hard to known optimum number of developers – Brook’s Law
used software to simulate this
empirical studies difficult – costly, OSS disparate, developers
comprehension not so good, implementation very little variation

Project Management Stage

I was thrilled to be invited to chair the Project Management Stage on behalf of the Agile Academy, and introduce two awesome talks by Paul Bannerman (Risk Implications of Software Project Organization Structures) and Steve Jenkins (Failing at Iteration Management – Analyst Style)