Title: The scoping problem

It is a cliché of software development that every project is understaffed with a deadline that is too tight, and maybe underresourced in terms of hardware and software too. I mean, have you ever worked on a project where there where too many programmers?

Jim McCarthy [McCarthy] suggests that every project is bounded by three factors: features, time and resources (i.e. developers) - these form the scope of the project: what do we need to do? how long have we got? who's doing it? Their product is a constant: if you want more features you must either increase the time allowed or the number of developers. Likewise, if you want a delivery sooner you must either reduce the number of features or increase the number of developers. But, as we know from Brooks' law [Brooks], increasing the number of developers makes a project later; unfortunately Brooks' law is not true in reverse: reducing the number of developers will also make the project later. Brooks' law can only be applied to a given point in time.

I'd like to add two observations of my own to McCarthy and Brooks:

I think from these observations flows the axiom: software projects are naturally over-committed in terms of time and resources.

If my observations have not convinced you, consider what would happen if you turned around at your next project meeting and said: "Well then chaps, looks like we've got four more weeks work and then we'll be finished a whole month early, well done." Would the reaction be:

I'd like to be bold and suggest Kelly's first law of project complexity:

You see, I think projects are bit like programs: they will expand to fill all the time and human resources available. This may be a generalised case of Brooks' law, in his case he adds more developers, now the project scope increases to cover training new developers too.

Every time a deadline is pushed back to allow more bugs to be fixed you are increasing the scope too. Even if the project introduced those bugs in the first place, fixing them was never on the feature list.

With a loose deadline there is a tendency to try some new beta OS, upgrade your compiler, or experiment with generic programming or many other things. Yes, these all have a place. I'm just saying you have to recognise the cost of all these things. Introduce a breathing space between projects to do these things. I don't want to argue for arbitrary deadlines either. A deadline picked from the sky and imposed is just as bad. Most projects have a rough deadline before they ever start, given this the team needs to look at the feature set and the resources, and balance all these things.

A word of caution: a developer joining a project at mid-point may not be aware of these discussions, it is important to explain the schedule to them and even adapt it after hearing their views. If all of this puts you in mind of short cycles you're right. Every time I work on a large project I am convinced that much of the project could be achieved with less code, complexity and effort.

This leads me to Kelly's second law of project complexity:

Take a look at your project, why is it so big? Would you write it the same if you were doing it again? Chances are you would not, you have learned from what you have done, and you would rewrite it using less complexity and less code.

Writing a program is teaching the machine how to do something. In doing this you come to understand the problem intimately and inevitably see better ways of doing it. So much complexity is unavoidable, but I suggest that much of it is avoidable if you merely reduce the scope of the project.

OK, how do we reduce the scope of the project? Well, firstly aim for minimalism [Henney] in design. Secondly, design extendable systems. Write a core system to which you can add code, your project may end up being big but it will still contain a microkernel like element which make it easier to get to grips with: embrace the words of John Vlissides [Vlissides] "A hallmark.... of good object oriented design is that you can modify and extend a system by adding code rather than hacking it."

Next: keep your project and your team focused. Each project iteration should have some Big Idea which you can drive towards. This is good for morale too and you always know when you have arrived.

And try this on your management: for every day the project is finished early, everyone on the team gets half a day's paid holiday.