Hi, Hi, > > My biggest concern about supporting friend classes is the ability to > access non-intentional to be accessed code outside of the original class's > knowledge. This by itself is very dangerous. >
Just to clarify, the original class explicitly declares friends of itself. There is no opportunity for friend classes to access protected members without the original class knowing about it. It is a specific modeling strategy used to improve encapsulation of behavior at the cost of tight coupling between private collaborators. In addition, other properties of friendship protect the original class from unwarranted access. The common expression of these properties is something like: "Just because I grant you friendship access to me doesn’t automatically grant your kids access to me, doesn’t automatically grant your friends access to me, and doesn’t automatically grant me access to you." In essense, it is actually a very-safe feature and while it can be abused, surely, I wouldn't go as far to call it very dangerous. I have seen some pretty heinous usage of traits, but I won't condemn the feature as dangerous! :) > I do see however package-private classes as a possibility (I actually have > a partially running patch for that) and allowing it to access protected > (not private) members is a second stage of my goal, so very similar to what > your suggested, with the exception that the original class defined what can > be accessed. > > Regards, > I see package-private classes as a separate modeling concern in the same domain of "private collaborators", personally. While package-private classes may be used to implement friend-like behaviors in a programming language, they do not clearly express the core intention of friendship as a tight-coupling between a limited set of collaborators (which may, in fact, be members of separate packages). To be clear, I don't want to pit two features in this domain of private collaborators against one another. I see them as 100% separate modeling concerns; each having its own strengths and wouldn't want to see something as useful as package-private classes be watered down to support friend-like behaviors because it would mean losing expression of the intent: specific targeted coupling of a few collaborators for better separation of concerns. Thanks for your thoughts! -- Dustin Wheeler | Software Developer NC State University m: mdwhe...@ncsu.edu | w: 5-9786 "If you don't know where you're going, it's easy to iteratively not get there."
