Most people in software are at least aware of Not Invented Here syndrome. We're in the middle of fighting with it right now, on the other side where we have some dependencies that are causing issues because we rely on some libraries for a very important piece of our software and the library has upgraded and if we use the upgrade we are, for all practical purposes, going to have to do a rewrite. So I was going to write this big long article in defense of NIH and why it makes sense to roll your own sometimes, but then I remembered that Joel on Software had tackled this already.
I reread the piece and it is basically saying the exact same thing I was going to say, so here it is. I can't say that I would go to Joel's extreme of "If it's a core business function -- do it yourself, no matter what.", but you need to seriously think about the cost of upgrading dependencies if the dependency is a core piece of what you are doing. It's not a matter of IF it will bite you in the ass, but WHEN and how hard.