I did my stint in college with modular software design and have always asked the question "At what stage does it get too modular?" Let me explain that question a bit more in depth - It is noce to have a modular and pluggable arachitecture but the more pluggable and modular you make it, the more overhead you have. The problem comes when you are putting int oo mcuh overhead and not really using it. This in my opinion is the call that the software architect is paid the big bucks to decide. The problem with making the call is that not all conditions in which the software will be used can be forseen. So when does a design get too modular?
I currently have a case where the software that we designed (notice the "we" now that the design turns out to be not optimal - just kidding) was not made pluggable in the data access part. Now it looks like under different conditions different dlls have to access different data modules. If I would have called for this part to be modular and pluggable in the first instance I would have gotten away with it. Now, I have to redesign the end componenet a bit. Not a good thing :( . Oh well, at least I know the set of conditions it occured under and the problems associated with it. A good lesson for me.
Oh yeah, so when does a design get too modular? Short answer - never! It just depends on the resources that you can spare.