I have run into many, many incidents like this. I have found that #3 and #4 can take as long as #5 depending on the system and proposed solution. It's is quite amazing at how undocumented some open source code really is.

For example, at one point, I took over a person's project. The project was half done and he took off from the position. So, I was left with pages and pages of undocumented source code. I had, in my mind, my own solution to the problem; but I thought, "this guy has already put in all this time." "I might as well build on his stuff." Unfortunately, I spent many long hours and long days reviewing his code, tweaking his code, hating his code, and building upon his code. I finally got the project done, but it was not an optimal (really, not even a good) solution. So, I ended up starting over from the ground-up after having finished the first iteration.

In the end, though, it really depends on the situation. A choice of using MySQL vs. MS-SQL vs. Oracle depends on a lot of things (money, do you need transactions, what will be connecting to the databases). Or, maybe it is groupware, do you go with Exchange or Novell or phpGroupWare? All that comes down to money. Sure, you could get phpGroupWare for free; but a lot of programming hours will have to go into it getting it to perform as well as having some specialists come out and set up Novell. Of course, if you want to do some customization of the system, phpGroupWare would be a good choice (maybe). But, if you want uber-customization, cooking your own would be good.

The essence of TIMTOWTDI.

Jeremy