CS 501: Software Engineering Fall 2000
Lecture 27
Software Engineering as Engineering
Administration
The Y2K Problem: Saving Memory
In 1967 memory cost $1 per byte The Air Force used single digit dates If 2-digit dates saved 1% of memory... savings over 20 years $16 to $24 million per gigabyte
The Y2K Problem: Saving Memory
By 1980s, memory was much cheaper, but 2-digit dates were standard. Why incur the cost of changing standards? 1970 The mortgage industry 1990 The Social Security Industry moved towards 4-digit dates On January 1, 2000 2-digit dates stopped working!
Where's the Problem?
A simple bug: dates of the form 19xx have been encoded xx A simple fix: find every occurrence of the bug modify the code recompile Where's the problem?
Find Every Occurrence ...
What computers do we use? data processing control embedded systems personal devices
What programs do they run? in-house development packages and libraries firmware, microcode, hardware Who wrote this program? Where is the source code?
Where's the Problem?
Computers fail everyday. What's special about this bug? What if they all fail at the same time?
What if we lose telephone, electricity, radio, etc.?
Traffic signals, elevators, The greatest worry was uncertainty.
Social Consequences
Worry creates its own problems: Wal-Mart forecast lower profits in Q1 2000 Legislation to limit law suits Opportunities for computer fraud and sabotage Trading partners
Organizational Procedures
Ostrich => do nothing => buy insurance
Bureaucratic => fill in forms that programs are compliant
Subcontract => hire Y2K specialists Do it yourself => in-house computing department
Y2K Validation
Request from Library of Congress to confirm that our code is Y2K compliant: Our code is fine .... but it depends on ... which depends on ... Yes. Our code is fine. Request from DARPA to confirm that our code is Y2K compliant: It's been validated by another part of the US government Thank you!
Technical Strategies
Replace noncompliant applications with compliant ones (e.g., new versions of packages) Repair noncompliant applications (e.g., in-house applications) Terminate noncompliant programs on an as-needed basis
Mask the data exchange between applications
Object code interception
New Bugs
If it's not broke don't fix it. 10 billion lines of code checked (often automatically)
10 million new bugs introduced accidentally
?? security holes, errors, etc. introduced accidentally or deliberately
Is all the Money Going to Y2K?
Y2K as a great excuse to have the computing budget increased: Upgrade the operating system
Replace the old package
Sell something to your customers What boss will turn turn a request for Y2K funds? What systems administrators will not install Y2K upgrades?
Profiteering
Buy gold, wood stoves, bottled water
Y2K specialists Pundits, consultants, writers Religious cranks
Final Thoughts on Y2K
We create computer systems that are more complex than our understanding of them: We over estimate our ability to validate systems
We under estimate our ability to adapt and respond
Software engineering usually thinks of systems as independent.
Will the long-term benefit of the Y2K problem be a greater understanding of how software systems interact with each other and with our social systems?
The Need for Software Engineering
Software as a product: => Awkward to use => Full of errors => No chance to try it out => No guarantees
Not much of a product
What is Engineering?
What is Engineering?
The profession of: ... creating cost-effective solutions ... ... to practical problems ... ... by applying scientific knowledge ... ... and established practices ... ... building things ... and taking responsibility for them!
Crafts, Science, Engineering
Science
Production
Commercial Craft
Professional Engineering
From: Shaw and Garlan
Crafts, Science, Engineering
algorithms data structures
Science Production compiler construction
software development methodologies Commercial
Professional Engineering
Craft
From: Shaw and Garlan
Software Engineering as Engineering?
Part craft -- part engineering Embryonic scientific basis Evolving body of knowledge
Too much flux for the apparatus of a profession (e.g., accreditation) Example: Texas and the ACM
The End
Good process leads to good software: the limits of heroic efforts
Minimize risks:
visible process function v. time v. cost
The importance of people
Requirements, requirements, requirements!