A fly on the wall perspective on the strange things project teams do to ensure that their Java project never manages to deliver anything useful.
The most important thing you have to do in order to crash and burn your Java project is to ensure that nobody who knows anything about OO design gets to work on the project. The easiest way to do this is to hire cheap, fresh out of school "Java wannabe" programmers and give the design job to someone who has never delivered an OO application. If that strategy fails and you get given a real designer make sure that some really junior programmers are assigned the task of "assisting with the design". If you do this really well your designer will be too busy helping the juniors to actually manage to do any design.
Make sure that you start the project with a really big team from day 1. After all, we all know that adding people to a late project makes it later, so we should start with a big team right from the start. A big team also mitigates the problem of losing people as the project progresses, so the best plan is to over-staff at the start, allowing natural attrition to reduce the team to the optimum size by the end of the project.
Starting with a really big team also helps prevent any real OO design from happening, after all with all those developers you have to start coding as early as possible. The earlier we can start coding the earlier the project will be finished. To add a bit of excitement to this mix make sure that the requirements are vague. The best way to do this is to create a really thick functional specification containing lots and lots of detail about how the application is supposed to work. The trick to making this work is to ensure that this 1,000 page "tree-killer" doesn’t contain any information about what the overall purpose of the application is or the business justification behind the application. After all that kind of information is too high level for mere programmers to be interested in.
If someone has the temerity to suggest that using Use Cases would be a better way of documenting the requirements, make sure that the Use Cases get written after the functional specification is complete. After all your users are familiar with functional specifications, so first the team must give them what they are familiar with and only then can they waste time on creating these new fangled Use Cases. For added fun make sure that everyone who gets involved in writing these Use Cases uses a different style. This is easier than it sounds, all you have to do is buy one of each Use Case book and make sure that each analyst is given an office next to their users who you have thoughtfully loaned one of your Use Case books to.
Having handled the requirements capture process you now have to turn your attention to the application architecture. If you already have a corporate application architecture then your best bet is to make a case that your project is special and needs to use a different application architecture. Select your best designers and programmers and give them a free hand to investigate alternatives. Make sure that they write up their recommendations in the form of a white paper. Once this is in place the rest of your team can ignore the corporate application architecture and do whatever they want to do. When the white paper is eventually ready, even if the recommendation is rejected it will be too late to change all of the code that has already been developed.
If you do not already have a corporate application architecture you are safe. Either roll your own infrastructure or select a bleeding edge vendor. Bleeding edge vendors are best because by the time your application is ready the infrastructure will be stable and state of the art. Make sure that you consider all possible vendors and have your complete evaluation process documented before you start so that nobody can say that the selection process was biased.
Rolling your own infrastructure can also be a winning strategy because these frameworks are not all that complex and you could always sell it to other companies when you are finished. It’s also a great way to use up your highly skilled talent on the project.
With requirements and architecture sorted you now need to turn your attention to the real developers who will be writing your application. Even though you have hired cheap you want to make your team believe that they are the best individual developers in the world. To do this encourage them to display their creativity and intelligence through the code that they write. Also, if any of them have the temerity to ask for a course you can squash that heresy immediately by implying that "if you can't get it out of a book, you must be stupid". This of course helps the entire team think that they are really brilliant and your attitude will prevent genuainely good developers from ever staying long enough to dispel the myth that you are creating.
To encourage developers to creatively express themselves through their code you need to ensure that you have an immensely detailed coding standard. Rather than stifle creativity by insisting on a single placement of braces, let each developer choose a construct and let them define that part of the coding standard. This avoids all those unproductive hours of arguing about the one true brace style and allows every developer to put their stamp on the coding standards. Do the same thing for comments, naming conventions etc., until you have a 200+ page coding standard. Circulate the resulting document throughout the company and set up a mailing list for discussions and feedback because meetings would be much too much of a drag on the project. Encourage flame wars on this mailing list as having an emotional outlet is healthy for the team.
Since code reviews are harmful to the creative expression of individuality by programmers avoid them if at all possible. If you have to do code reviews you must ignore all of the feedback from them. Even if there is a list of items requiring changes, postpone working on that code because it’s much more important that you add more new features to the application.
Remember all that stuff about "encapsulation" that OO books talk about? Don’t touch it. Private methods make it impossible to use inheritance. Ideally everything should be public so that it can be accessed in the most efficient manner. The other problem with encapsulation is that forces developers to write lots of inefficient little methods that are scattered across lots of different classes. To add a simple feature developers have to modify lots of different files and even with the best will in the world it’s easy to make a mistake when modifying lots of files. The best place to put all of the code is either behind the OK button (you only have to go to one place to see what is happening) or in stored procedures in the database (since these give optimal database performance).
The last thing you have to remember is that developers always over estimate the time they need to write the code. You need to question all estimates and make every developer justify why they need to take that long. A good put down line to use is "I’ve written more complex things in half the time." If any developer is unwilling to reduce their estimates find them a much simpler task that even they can do and give their work to a real programmer.
(C) 2001 Software Craftsmanship Inc., all rights reserved.