A common problem is that few development organizations take as much care in the design of their error handling as they do in the functional handling. Without this care, one result is that method signatures have too many caught exception declarations. This leads to poor exception handling in the caller and so on up the call stack. A good exception design -- for example Spring's -- leads to very few exception declarations on the method. I think that Spring's next step of hiding the exceptions is a mistake, but their code is far more used than mine so who am I to argue!
It is worse as you also have to consider the false negatives. Assume (for ease of calculation) that 1% of the publication are terrorists. Within a population of 1M there are 10,000 terrorists. Within that, 2% of terrorists will not be recognized. 200 terrorists are allowed to fly. Boom! It really doesn't matter what the real ratio of terrorists to non-terrorists is. What matters is the human costs of a false negative.
I highly recommend working on simply porting an existing tool to a new environment. You will be providing a useful service and learn piecemeal about an implementation style and technique. The more you port the more implementation dos and don't you will learn.
Slashdot Mirror
A common problem is that few development organizations take as much care in the design of their error handling as they do in the functional handling. Without this care, one result is that method signatures have too many caught exception declarations. This leads to poor exception handling in the caller and so on up the call stack. A good exception design -- for example Spring's -- leads to very few exception declarations on the method. I think that Spring's next step of hiding the exceptions is a mistake, but their code is far more used than mine so who am I to argue!
It is worse as you also have to consider the false negatives. Assume (for ease of calculation) that 1% of the publication are terrorists. Within a population of 1M there are 10,000 terrorists. Within that, 2% of terrorists will not be recognized. 200 terrorists are allowed to fly. Boom! It really doesn't matter what the real ratio of terrorists to non-terrorists is. What matters is the human costs of a false negative.
I highly recommend working on simply porting an existing tool to a new environment. You will be providing a useful service and learn piecemeal about an implementation style and technique. The more you port the more implementation dos and don't you will learn.