Monday, January 24, 2011

Extreme Programming Ch 4-6 and Design of Future Things Ch 2

Extreme Programming
By: Ron Jeffries; Ann Anderson; Chet Hendrickson
Publisher: Addison-Wesley Professional
Pub. Date: October 16, 2000

Chapter 4 : User Stories
The XP programming methodology advocates the user of "user stories" that are made by the customer to introduce requirements and functionality in a narrative format. These user stories are made in meetings with the programmers and each story will have only a small set of features written down on a note card. Customers are encouraged to rip up their cards to act as kind of an ice breaker where the team isn't afraid to dispose of bad ideas. Programmers are forbidden from writing stories or asking influencing the customer with ideas about the implementation of the stories. A typical project will be composed of many stories; The book gives an estimate of 60 to 120.



Chapter 5 : Acceptance Tests
Since customers have the right to see progress on the project and because progress is measured by tests of the system, XP advocates frequent tests, particularly unit testing with test automation, on each iteration. Frequent acceptance testing reduces bugs and speeds up the project compared to if testing was done right before product release. Furthermore, the customer can evaluate the testing results and provide feedback.

Chapter 6 : Story Estimation
Since project requirements and features take the form of user stories, programmers must learn how to estimate how long it will take them to implement a story. Estimates are made on a point basis (rather than time) where each programmer makes the estimate by comparing similar stories/features, keeping a "perfect engineering week" frame of reference, and by doing "spike solution" for stories that are difficult to make a comparison to. A spike solution is coming up with a general idea or set of steps for solving a new problem. The spike solution isn't thorough but just a rough estimation.

The Design of Future Things

Design of Future Things:
By: Donald Norman
Publisher: New York Basic Book
Pub. Date: November 2007

Chapter 2 : The Psychology of People and Machines

Norman begins the chapter by giving a brief sets of example scenarios of machine to human communication and claimed that machines will soon be communicating autonomously
or semi-autonomously. Norman highlights the differences between machine and human thinking by pointing out that machines, while fast, lack the kind of massively parallel computation and pattern recognition that humans have. Norman also says that he demands of the real world create a need for autonomous agents to tackle problems. Norman reviews a points from his previous book, "Emotional Design" that outline the three levels of human processing:

Visceral - Basic, automatic, subconscious
Behavioral - learned skills, mostly subconscious
Reflective - Self-aware higher level thinking

After reviewing these terms, Norman proposes that machines should be imbued with these levels of thinking and feeling ("emotions") to enhance performance and feedback between human computer interactions. This would create the symbiotic effect that was described in the first chapter.

Norman also points out that machines have problems with goal formulation and deciding how to bridge those goals. Furthermore, machines measure things differently and often can't navigate the world as well.

Finally Norman discusses the issue of "common ground" that informs communication beyond its content. Norman claims that machines need background of protocols with humans to communicate. Even more than that, though, is the need for common perception of the context in order to establish common goals and problem definitions based on observation. Norman describes this as "bridging the super-gulf."

Discussion:

Although using cards to write down story scenarios sounds light and digestible, which is nice, I have my doubts on how complete a system's features and requirements can be elicited especially if the customer is the only one writing the user stories. I do like the idea of unit testing early and often. Looking back at past job experiences I know I could have benefited from that kind of rigorous testing. Errors are much easier to track when caught early because the code is fresh in your mind and there's less interacting factors to take into consideration. The fact that user stories are made in bit-size chunks and that time estimation of the project is done by evaluating these pieces, I'm sure that XP does a wonderful job of time estimation overall.

While Norman's advocacy of emotional machinery is unique and an insightful analysis in terms of its links with communication and synergistic processes, his descriptions fall flat in terms of providing means to implement these kinds of reactions. Ultimately his observations of the limitations of machines are perhaps the easiest to digest and correspond to current areas of research in parallel computing, neural networks, and, in particular, AI. Normans discussion hit on AI's fields of computer vision (and other sensing), robotics (mobility), mapping of percept sequences to action states that reach a goal state and how to form a goal based rational agent. Although Norman's perceptions mirror AI's fields of study, his observations on needing a common ground to communicate and formulate problems with machines is insightful.

No comments:

Post a Comment