I was asked on my prior post what process I follow once I have decided to redesign an existing application. This is not an easy question for me to answer partly because the process is dependent on the application being redesigned and also because my general approach to the problem evolves with each new project - whether it is a redesign or an entirely new application. However, some elements of the process are always present.

First off, I need to understand the existing application. Sometimes I was the person who wrote the initial application but typically that is not the case. Nor do I usually have any real documentation on what the application does and sometimes I don’t even have the source code (such as when we are tasked with writing a new generation of software for a government system and the initial contractor doesn’t even exist anymore). While learning to understand the application I make sure to document all of the functionality that I can in both use cases and the UML. This helps me get a better grasp on what I will need to do, plus help me identify any existing holes in the current system.

Once the current systems functionality is documented and understood I can begin to identify the objects that make up the system. Via my prior documentation I can typically see most, if not all, of the inter-object relationships as well. Once again, I return to writing documents and crafting UML diagrams.

As you might imagine, this is pretty boring stuff. But it needs to be done else nobody working on the final product will have a clue what really needs to be done and we won’t have any real requirements to verify against.

Now that I have a set of objects defined I work on creating a data model that will support my objects. The data model doesn’t necessarily match up with my objects on a object::property <–> table::column basis but most of my object::property values do need to be represented somewhere in the data model.

Up to this point I probably haven’t even settled on a data source nor an implementation language. At times the customer might have already specified both - but sometimes we have the choice of actually picking what we think will be the best tools for the job.

At some point we do have to actually build the system though so once all the analysis and modeling are “done” some code will get written. I say “done” because the models will all change many times as I come to a better understanding of where certain methods should be within the system and as I gain a stronger understanding of the system as a whole.

It is also important to remember that the customer (whomever that is, be it your boss or an actual outside entity) needs to stay involved throughout the process to make sure you are actually giving them what they want and not necessarily what they said they want.

Communication is a tricky thing and this is probably the most important lesson I have learned. Until the customer actually sees what they asked for in a system they don’t necessarily always know what they really wanted. It is important to make many small point builds on a regular basis that the customer can see/touch. By doing this they can give more consistent and frequent feedback. I for one would rather know as soon as we do something that didn’t match the customers vision than weeks or months down the line when a change may be exceptionally difficult/expensive.

I know this doesn’t necessarily explain my process but it is about as well as I can do in a short hopefully readable post. Later this weekend or early next week I will try to explain what I am going through with ProTask - an application I inherited and already redesigned once and am in the process of redoing once again.

My degree is in “Computer Science and Software Development” which is about as close as my university could come to claiming it offered a true Software Engineering degree. Needless to say, I came into my profession without the best background as almost everything I learned at school was self taught. Because of this my work has taken giant steps since I entered the industry 5 years ago. However, this means that some of my earlier work has plenty of room for improvement.

Finally, here are a couple of books that I have read that have been both useful, informative, and that I recommend.