Software CSE403: Engineering
Reid Holmes Winter 2009
COMPLEXITY
NSB REDUX
Confusions
over accidental complexity Emergence of OOP "The best way to address the complexity of software is to not build it at all" Several NSB responses highlight software reuse As a means to reduce complexity As a means to improve productivity As a means to increase reliability
SOFTWARE REUSE
Domain-specific
component markets Populated by carefully created reusable components New features are added by dropping in components Accounts for 53% of reuse at NASA [Selby 2005] Three main impediments: High up-front cost [Gaffney 1992, ICSE] Library scaling problem [Biggerstaff 1994, ICSR] Architectural mismatch [Garlan et. al. 1995, IEEE Software]
ECONOMICS
Budgets
are drawn up annually Heavy emphasis on the current quarter Reusable software is: ~Twice as expensive [Gaffney 1992, ICSE] ~Three times as expensive [Brooks 1975] Requires careful forethought to determine what software will be reused and whether any savings outweigh extra costs What is the benefit to the customer?
LIBRARY SCALING
Two
extremes: Large, feature-laden, components Small, simple, components Adapting large components to a system can be difficult The effort of adapting a small component might outweigh any benefits of reuse in the first place
ARCHITECTURAL MISMATCH
Even
reusable code makes some assumptions about how it should be reused; these assumptions are often implict Explicit assumptions are often easy to identify: Programming language Libraries & frameworks Implicit assumptions are harder to spot: Topology assumptions Protocols of use Implicit assumptions are often not documented because the original developer may not have considered them constraints
AN ALTERNATIVE REUSE APPROACH
Bug: UltiGPX should I wish UltiGPX showed me visualize elevation how my elevation has changes in trackschanged... using a profile view
UltiGPX
8
MOTIVATION
Azureus UltiGPX
1) Enact
9
MANUAL Plan
2) REUSE APPROACH
Easy to get discouraged Difficult to modify earlier decisions Easy to attempt infeasible tasks Piecemeal nature diverts attention from core technical difficulties
10
PRAGMATIC REUSE
White-box
reuse Code Scavenging [Krueger 1992, ACM Computing Surveys] Ad hoc nature increases risk of bad decisions Adaptation expensive and overwhelming Industrially effective Effective reuse approach [Frakes 1995, CACM] Common risk-aversion practice [Cordy 2003, IWPSE] Replicate & specialize [Kapser & Godfrey, 2006, WCRE]
11
PRAGMATIC REUSE PROCESS
Identify starting point Plan task / triage dependencies Enact plan Evaluate reused code
Abort task Re-implement feature
12
PLANNING A TASK
4
main kinds of decisions: Common Accept Reject Remap
SpeedGraphic.drawChart()
Legend
O Method or Field ! Call or Reference
13
TRANSITIVE IMPLICATIONS
Accepted Code
Rejected Code
Remapped Code
Common Code
14
ENACTMENT PROCESS
Existing Code
Accepted Code Rejected Code Remapped Code
Developer's System
1) Extraction
2) Integration
15
H-1: TASK TIME
Task Time 30 Time (min) 10 15 20 25 35 40
5
Supported Reuse Manual Reuse Task Su...