It's interesting that a major point is basically to trust cryptology standards. I have seen this statement echoed by various crypto guys around the web.
There is a real convenient tendency when programming/desigining to perhaps exagerrate ones own knowledge, or at the least try to build something from what little knowledge you have. In many areas of CS, this is probably a good thing. Afterall, sometimes you learn best by doing. However, having seen a number of homebrew suggestions around the place and having seen them shot down, I remain convinced that crypto is complex and difficult enough that we must leave it to the experts. Unless one can formally and rigourously describe the strenght of their scheme (i.e., in the sort of mathematical language that give me a migraine), homebrew solutions will never cut it.
Nevertheless, people are constantly deluding themselves into believing their "unique" solutions are better than a systems devised by uber-geeks and techno-spooks