Wednesday, March 30, 2011

Mythical Man Month Ch 10-12

Mythical Man-Month, The: Essays on Software Engineering, Anniversary Edition
By: Frederick P. Brooks Jr.
Publisher: Addison-Wesley Professional
Pub. Date: August 02, 1995

Chapter 10 : The Documentary Hypothesis
Brooks claims that there's a small set of critical documents that
manager are necessary while other considerations may be mere distractions.
The critical "documents" (or critical points to cover in documentation) are:
Objectives
Specifications
Schedule
Budget
Organization Chart
Space Allocations
Estimate, Forecast, Prices

Brooks expounds on the last point by saying that these three variables feed
into each other and create a negative or positive "spiral." The flow of the spiral depends on whether the prices are below or above the budget estimates. Brooks notes that the negative spiral requires a discipline among all members of the team to trim down the project and get the most performance in order to ultimately succeed.

Brooks also takes a look at the necessary documentation in an university department and finds that they correspond to the average computer project in that the main components include objectives, product specifications, time allocations, money allocations, space allocations, and people allocations; or in other words, the "what, when how much, where, and who." These same considerations also carry over into software projects.

Brooks cites that the act of writing formal documentations reveal the holes in system design and solidify requirements for the project. Secondly, the documentation helps fulfill the roles of communication required of the manager and "lighten his load." Finally the documentation sets down a check list to monitor progress.

The task of a manager, claims Brooks, is to develop a plan and realize it.

Chapter 11 : Plan to Throw One Away

Brooks, realizing that the first implementation of a complex system is bound to have bugs and not adequately compensate for the difficulties inherent in such complex systems or be able to scale up for large scale use. Brooks purports that the first implementation of a system should be thrown away and that throwing the first away is inevitable, so the team/manager should plan to do so anyway.

Changing requirements of the client and changing technology is a fact of life, so management should embrace change and be flexible, but only within a limit, otherwise, the project would never reach an end. Management is to embrace the fact of continuous change by planning to throw one away, documenting changes to reduce errors, and careful tracking of the schedule and speed of implementation.

Brooks also notes that the entire company organization should be flexible to allow for change and to allow for technical and managerial people to both perform high level work and get their hands dirty in lower level implementation work so that their skills are shared and talents can come to forbearance. To do this, managers need technical refreshers, technical people need management training and problems on either domain should be communicated to the other.

From there Brooks has a word on system maintenance and notes that bug elimination usually proceeds as a "two steps forward, one step back" process since local bug fixes tend to introduce bugs at the system level. As system maintenance proceed, a kind of system entropy is introduced where bug fixes follow a pattern of "one step forward and one step back"

Chapter 12 : Sharp Tools



Brooks cites that one of the main problems on programming projects is communication, and as such, "individualized tools hamper rather than aid communication." Furthermore, technology changes across machines so tools don't last long. However, there is still a need for specialized tools for specific tasks. The problem is complex enough that Brooks proposes that a specific tool specialist should be employed to manage and update tools for the entire team. Brooks also claims that a separate computer facility be available and standardized languages, OS, utilities, debugging aids, test case generators and text processing system be used.

Access to a common computing resource by multiple small teams should be done by giving them a longer dedicated session that lends to productivity rather than splitting the resource and maximizing access by using smaller sessions. Brooks also advocates usage of graveyard shifts to maximize productivity. Brooks continues to cover simulators, compiler and assembler "vehicles" (machines), program libraries, program tools, documentation systems, and performance simulators.
Key points included providing programmer copies that are separated from formal releases using high level tools, word processing, and simulators to enhance productivity, communication, and reliability, respectively.

Discussion :

Despite the overall adherence to heavy formal documentation in previous chapters, Brooks switches gears and claims that documentation, while it needs to be thorough, can be vastly simplified into a small set of categories that answer the basic questions that anyone would have for a system. From personal experience, I find this to be true. Setting out the what, when, where, how much, et cetera really does communicate to both the team and overseers and gives a clearer view of the project ahead that allows things to proceed more smoothly both from the managers perspective and overall-progress-made perspective.

I've heard the famous phrase "always plan to throw one away, you will anyway" from multiple professors and other sources, and I never realized where the idea originated from. Once again, personal experience bears witness to the truth of the phrase despite my natural repulsion from the idea. I will say that I think that the first implementation to be thrown away should be done rather quickly as is promoted in agile development methodologies so that not too much time or effort is wasted and so that the team members don't grow too attached to the code that is going to be thrown away. The idea is to learn how to approach the problem and discover problems early and I think that should have been emphasized.

Finally, the chapter on "Sharp Tools" fell a little flat to me. Perhaps its the age of the text. I think there's been enough improvement in some of the lower level computer systems and enough progress on standardization that a lot of his concerns about tools and consistency invalidates the need for a dedicated tools specialist. Usually the software that's going to be used on a project is discovered early on and in such a way that the target system is developed adequately for and usually the software is easily ported to other systems.

No comments:

Post a Comment