Wednesday, June 27, 2007

10 Reasons Big Projects Suck

Have you ever noticed that big projects inevitably get a bad rap? Here are 10 reasons why:

  1. Let's assume for a moment that there's one bug for every 100 lines of code. If a big project has 10 times as much code as a small project, it has 10 times as many bugs. In reality, because big projects are harder to understand and intrinsically harder to change quickly, it probably has more than 10 times as many bugs.

  2. If a big project implements some feature A, there is bound to be some bug in it. That proves that the big project is buggy. Furthermore, inevitably, the feature isn't exactly what you need. That means it's inflexible.

  3. If, on the other hand, the smaller project doesn't implement feature A, it can't possibly have the same bug the big project has. Hence, it's not buggy. Furthermore, since you'll need to implement feature A yourself, you'll probably implement exactly what you need. That means it's more flexible.

  4. Furthermore, there are a lot of people who don't even want feature A. That proves that the big project is bloated.

  5. If a developer is a member of a big project, he is probably already using it in production, and he doesn't much care what some young, know-it-all kid says about his code. Ever wonder why Microsoft doesn't seem to care when people criticize it? They're too busy making money!

  6. However, if a developer is a member of a small project, he can afford to make fun of the big project. No one knows who he is, so they surely can't insult his work. He has security by obscurity!

  7. Furthermore, since so many people have worked on the large project, he can insult it vehemently without feeling morally responsible for insulting another person's hard work. It's like a shoplifter who shoplifts small items from large stores thinking the large store is too big to care.

  8. Let's suppose 1 out of every 10 projects succeeds. 9 of those projects will make claims that turn out to be false. However, since they don't succeed, no one remembers. However, the 10th project will make claims that turn out to be true. It has instant credibility. Hence, it is free to make claims, and many people won't even bother to verify or question those claims...at least until it becomes a big project and people start realizing that 9 out of 10 of its claims are actually false.

  9. If you only need to implement 1 feature, you can do so in code that is very simple and direct. Now, if you need to implement 10 features, there is bound to be some duplication. Hence, you can either a) live with the duplication, or b) refactor. If you live with the duplication, your project will be plagued with bugs that need to be fixed in multiple places. (Don't repeat yourself!) However, if you refactor the code, you'll end up with code that is (necessarily) more complex than when you only needed to implement 1 feature. Younger coders may not even be able to understand the code at all. Hence, they'll just call it stupid, bloated, and overly complicated.

  10. If a project is successful, it'll make it into production. Furthermore, people will need new features in the product. In implementing those new features, it may be necessary to refactor. When you refactor, you may need to decide whether to a) keep the existing API, b) re-write the API, c) create a compatibility layer. If you keep the existing API, you'll have to somehow "tack on" the additional functionality within that API. This may result in a hideous, unintuitive API. If you rewrite the API, you'll break everyone's code. If you provide a compatibility layer, you'll end up with twice as many APIs you need to support. Hence, implementing new features is the fastest way to end up with legacy cruft!

If you know more reasons why big projects suck, post them below! :-D

6 comments:

sigfpe said...

With a small project you can be on to something new in no time. With a big project you have a boring albatross around your neck for a big piece of your life...

Anonymous said...

Large projects suck because large project documentation sucks. It's easy to document a single program within the program itself, but as the number of programs increases, the number of conditional interactions increases, which makes reading the documentation more difficult than going to the original source code. Perhaps we in the Python community should come up with some doc standards that would make documentation work as well as Python.

Adam U. said...

Big projects suck because more time is spent maintaining old code rather than writing new code. If the documentation sucks (assuming there is any to begin with) you likely spend 99% of your time reading code and the remaining 1% writing patches or sitting in meetings talking about writing patches. :-)

Conal said...

A big project means that the problem model is complicated and could either be (a) simplified & generalized through a deeper understanding, or (b) teased apart into genuinely orthogonal, composable sub-problems. While the problem description is complex, the parts will have unknown interactions and gaps, leading to more feature requests & more complexity. So hire a genius to simplify & generalize the problem, and then form a small project to solve this simpler & more important problem.

"Simplicity is the most difficult thing to secure in this world; it is the last limit of experience and the last effort of genius." -George Sand

"Any intelligent fool can make things bigger and more complex ... it takes a touch of genius -- and a lot of courage -- to move in the opposite direction." -Albert Einstein

john said...

There is a corollary to why big projects suck. Big projects are usually associated with big companies. As a consequence they usually have operations and control structures, eg. meetings and status reports, that are required to justify the salary of some VP that earns 10x what you do and has no clue how the friggin program works either technically or economically.

Jessta said...

Big projects(eg. Firefox openoffice) make it very difficult for users to excercise the freedom to modify and improve the software given to them by the GNU GPL. The reason for this is that an individual would have to spend many months studying the firefox code before having any chance to make any meaningful improvements.