This is my first post to the Boost distribution list, so please feel free to speak up if you find any part of this message to be outside the normal decorum. I am curious if there is support for what I'm calling a "dual_state" template class. Anyone familiar with Perl will instantly recognize the parallel. The idea is to attach a "statedness" value to arbitrary objects, allowing us to consider them defined or undefined. Newly constructed dual_state objects are undefined. For example, // -- begin template < class T > class dual_state { ... }; dual_state<int> a; a.defined(); // FALSE defined(a); // same thing // -- end We can "define" a dual_state object by assigning a value to it, either as a base type, T, or as another dual_state object: // -- begin // assign an int to a a = 5; a.defined() // TRUE // assign another dual_state<int> to a dual_state<int> b; b = a; b.defined(); // TRUE // -- end We can undefine a defined dual_state object: // -- begin a.undefine(); undefine(b); // -- end In addition, we can utilize an instance of another class, the "undefined_object", to interrogate or set the defined property of a dual_state object: // -- begin const class undefined_object {} undef; a = undef; a.defined(); // FALSE a == undef; // same thing // -- end The dual_state class provides a conversion to T. This conversion will, for an undefined dual_state object, return default T objects generated using the expression "T()". As an aside, it seems Boost.value_init may provide a more general solution that is worth considering. There is no conversion to T& because some compilers are utterly confused by such conversions. To make dual_state compatible with large or uncopyable objects, two explicit conversions are provided: // --begin template < class T > class dual_state { ... const T& value() const; T& define(); }; // -- end The value() member is straightforward. As for define(), it is necessary to first define the dual_state object before assigning it a value. The operator= member normally takes care of this step, but when dealing with the T reference directly we must explicitly use define(). An example using define() is shown below: // -- begin dual_state<int> x; cin >> x.define(); // -- end In Perl, the defined/undefined concept is tremendously useful, so I would expect similar utility in C++. I have not given any thought to what a dual_state<bool> is or how it is related to Boost.tribool. Thanks to everyone who took the time to read this whole post. I am open to all suggestions, including a better name -- "dual_state" is quite incongruent, but I can't think of anything better. Here are some additional poor suggestions: bistate, managed, switched, stated, definable. -- Andrew M. Jost