This chapter is from the book
This chapter is from the book
This chapter is from the book
PEOPLE
The three most significant people on your project are the executive sponsor, project manager, and technical leader. The absence of any one of them can cause almost any project to fail. The executive sponsor is responsible for getting the project the "nutrition" and support it needs. The project manager is responsible for the quality of communication between the teams. The technical leader is responsible for ensuring that a design is used that both satisfies requirements and can be built by the team.
Some other people who may influence the success or failure of any OO project are the technical staff and the users. You also will want to consider how personality types could affect the overall outcome of the project.
Executive Sponsor
The executive sponsor is a critical success factor, according to many project leaders I speak with. In many cases, the sponsoring executive is not in the project developer's immediate line of command. Sometimes, the sponsoring executive is in the user community (see, for example, K.L.'s Eyewitness Account in Chapter 5). The point is, find that supporting executive, wherever he or she might be.
Do not, however, invent all sorts of productivity gains for object technology in your effort to gain executive sponsorship. You are unlikely to achieve such gains on your first project unless you are deliberate in getting highly experienced OO developers to execute your project, and false productivity claims may instead put you in an awkward position during the project.
Try not to make any claims for the technology that cannot be demonstrated on your project. Instead, examine your organization's situation and understand where it is suffering. If you can identify one or more of the probable benefits of object technology that will help with that, then it will be reasonable for your executives to consider the technology.
Support for object technology comes from one of three places:
The executive team. They know where they are troubled, and have looked at a number of options. They may be ready to try object technology as the next logical alternative.
The users. The user community often funds or affects funding. They may have seen OO projects done with close user interaction, perhaps even designed by users.
The programmers. They may have gotten the latest C compiler, which happens to be C++, so there is no way to stop them from working with objects. Perhaps the organization is dominated by the programmers, and when they decide to try object technology, the management follows.
Whichever of these is your source of support, try to get the agreement of someone who will support you when you get stuck and need to make changes in your proposed project. You will have peace of mind if you know your sponsor is behind you at some potentially embarrassing moment in the project (see Project Winifred in Chapter 2).
Project Manager
All projects rely on management skills. The project manager must be able to judge the expertise present, how and when to improve it, and how to scale back the project if the expertise is not sufficient. A good project manager, even one with no OO experience, will be able to adjust for the few different considerations that object orientation creates. If I had to choose between a good manager with no OO background and an OO-experienced but inept manager, it would be no contest; I would choose the good manager. The project manager's lack of OO experience can be filled by the technical lead; but the absence of management skills cannot be covered by anyone else. The project managers I interviewed were unanimous that objects change the project manager's life relatively little, while changing the developers' lives significantly.
Technical Lead
If your project manager has no prior OO experience, make sure you have a technical lead with good communication skills. The technical lead can let the project manager know when the project's needs are unique to objects. (This book should also help.) If your project manager is not comfortable with "object-speak," and the technical lead is not helpful and communicative, then you have a problem that needs to be fixed.
The qualities and skills of your technical lead are related to the type of project you are undertaking: SWAT, Investigative, Production, or Full-Commit. SWAT and Investigative do not require that the technical lead have OO experience, but Production and Full-Commit do.
Technical Staff
Technical staff includes your experts, your regular staff, and the mentors. It is important to distinguish between experts and mentors. The responsibility of the design expert is to make sure the system gets built, and to a sufficient standard of quality. His or her key characteristic is technical proficiency, with communication skills secondary. The responsibility of the mentor is to improve the staff's ability with objects. The mentor must know how to practice safe design and teach it. His or her key characteristic is the ability to communicate clearly, with technical proficiency necessary but secondary.
More object projects are starting up than there are experts to lead them, and even large consulting houses are having difficulty finding experienced OO developers. This means two things to you. The first is that after you train novices for a year, they may quit and become self-declared experts. The second is that the expert you hire may be someone else's former one-year novice.
The protection against the novice–expert is simply to check their references and experience. I advocate that programmers carry portfolios with samples of their programming and design work. These samples can tell your interview team more about a programmer's thought processes than hours of talking. If a programmer does not have one, consider asking her or him to do some small task as a sample even if you have to pay to get it done. After all, you are thinking of paying $100,000 for this person over some time period.
To prevent people from quitting, you will have to examine the way you educate and use staff. The market value of your designers goes up as you train them, so either pay them more to keep them, or take a certain attrition rate into account as part of your project's survival. Remember, it costs at least half a year's salary to introduce a new employee to the organization. I suggest paying the good ones enough to stay with you.
Users
Put "real users" on your requirements and design teams. Many intermediaries incorrectly claim to know how the users work and what they want. Read Jon Marshall's Eyewitness Account (see Chapter 5) for an example of how differently people think compared to how others think they think. In that project, in Project Winifred (see Chapter 2), and in the Smalltalk projects in K.L.'s Eyewitness Account (see Chapter 4), success involved letting the user community create the system, and in one case, actually design part of it. On Project Manfred (see Chapter 2), by contrast, one of the team's criticisms was that they had a changing stream of users who were only superficially involved in the process. User Involvement, in Chapter 5, reviews the topic of links between developers and users.
Consider assigning an experienced and articulate user to the project full-time. This person will become familiar with the capabilities of the programming technology, and will carry that knowledge back to the users. Experienced users will pay for themselves over and over again by checking assumptions and smoothing the transition to the new system.
At the very least, have a few users available periodically, several hours each week, to discuss ideas and to make their work understood. Some projects gain an understanding of the users' priorities and obstacles by having a few designers work within the user community for a few days and by videotaping the process. A rare and farsighted project will hire someone to work with, videotape, and study the users at work for several weeks to identify where the process breaks down. These projects find the places where the most good can be done, instead of where just some good can be done. The bottom line is, everything you do to involve users strengthens your project.
Personality Types
Is there one personality type that is better suited to object technology than all others? This question is relevant to your project because it appears the answer may be yes. That answer does not come from any detailed study but from informal conversations I have had with other designers, consultants, and trainers. What could cause this? Here is an argument built from plausibility, backed by personal observation.
The ideal object designer/programmer thinks abstractly, deals well with uncertainty, and communicates reasonably, in contrast to the quiet, detail-oriented programmer stereotype. Good designers rely on the ability to look at things in the world, at words and concepts, and to extract from them properties that they share, that can be named, and that can be classified into a hierarchy. An object called a strategy is by no means easy to think up, and yet it becomes one of the most powerful elements in the system. What are the properties that all your user-interface screens have in common? The quality of the answer depends on the ability to identify commonalities and imagine various futures. Thus, the quality of OO design hinges on abstract thinking.
Most OO projects are run according to an iterative scheduling process, iterations, 123 often with aggressive changes across iterations (see Chapter 5). This means the requirements, the internal design, and the user interface may be in flux throughout much of the project. Programmers used to working from fixed specs may not feel comfortable in this environment. It is actually not object technology, but the iterative process that causes this discomfort.
Finally, there is an additional communications burden that comes with object technology. Because of the intensive use of interfaces, the designer is not going to be able to sit alone in a cubicle and crank out code from a spec. The design of the interface is crucial to the success of the class, and must be examined by other designers.
Although I am describing the ideal designer, most of the good object designers I know fit into this category. Also, I gave their characteristics in priority order. Abstract thinking is most important, dealing with uncertainty is second, and communication skill is the most dispensable. What happens to the people who do not fit this profile? The following true story illustrates. Note in this story that program designers benefit from being abstract thinkers, but requirements analysts need to be thorough.
Kim and Megan were teammates, responsible for the requirements and preliminary design of the business classes. Kim was particularly detail-oriented. She was proficient when it came time to go over the requirements document to see whether we had researched all the system interfaces, whether our drawings were consistent, and whether we had neglected any issue we had agreed to cover. She was also excellent at working out our schedule. But we could not engage her in a conversation about the general nature of legal agreements or policies. After months of coaching, her designs still looked either like the screen specification or the description of the relational tables, and her code still looked like COBOL. She badly wanted to do object design, but she was not getting there.
Megan, on the other hand, had no patience for paper-and-pencil work, but she was quick to sort out the differences between insurance policies and general contracts. However, she repeatedly overlooked requirements and interfaces, and was sloppy in her specifications.
Over the course of the project, we let Megan take on the internal design role. We could not afford to have her overlook critical interfaces and we could make use of her subtle appreciation for the differences between classes. So she was told the requirements, and spent her time designing classes and, eventually, frameworks.
Kim became indispensable on the requirements team. She made sure that all issues were covered with the users or the executives, that all interfaces were defined, that the schedule was updated, and that the design documentation was in order. Periodically, she asked to design some business classes, which we permitted since major flaws in her designs would be caught and corrected. The advantage to us was that she stayed highly motivated and paid her way fully in her other work.
The ideal OO designer thinks abstractly, deals with uncertainty, communicates with colleagues.