> Yes, I agree it's entirely possible that people will in fact do stupid > things with "private" variables. History has shown that's true. The > only thing I can ask here is that all the available technical > documentation for the private variable interfaces explains _clearly_ > why creating a private variable, though it looks "easy," is in fact a > Very Bad Thing, and suggesting strongly that well-defined public ones > are better.
So suppose they need to add a way to configure new feature X, where X is likely to become commonplace. Suppose there's no framework-level (aka public) support for feature X yet. Of course, it would be ideal if they worked with the OpenSolaris community to add framework-level support for feature X, but I think that's quite unlikely. I'd rather see them use a private property than invent some other hacky one-off mechanism. -- meem
