I too am disappointed in D because I hoped it would be an improved C++. But in fact it's an entirely new programming language, so many advantages of C++ are left out.

I'd like to see a programming language that is
-C++
-without headers, and isntead an alternative way to link source files together
-with you never needing to duplicate function declarations/definitions (this is part of the "remove header" plan)
-with binary literals (0b01011101001)
-with nested functions
-with backwards compatibility to C removed
-with an improved standard library, including standard libraries for simple graphics, GUI, audio, network, file management, mouse, realtime keyboard input, printing, terminal and multithreading
-with easier support for making vectors pointers to a mix of parent and child classes with virtual functions, like pointer containers that automatically destroy objects, note that pointer containers aren't in the standard at the moment
-with a programmer NEVER needing to use "new" and "delete", with all situations that could require the use of those having a replacement that auto-deletes (for example the vectors and pointer containers)
-with binary literals
-with a MUCH simpeler form to use iterators to elements of containers, I'm sorry but I dislike a variable type with a name like std::vector<int>::iterator, there has to be a better way to handle this, if needed make containers and iterators part of the language itself, for example the type int+ could be an iterator to containers of ints
-with more operators that can be overloaded
-with the ability to use the ` character instead of the " character for strings (I took this idea from D)
-with array literals
-without the inline keyword, it's not needed anymore
-with certain undefined things defined, like some special operator orderings there are and cases with the % operator
-with 64-bit integers, 128-bit integers, 128 floating point numbers (variable types that are defined to have at least that amount of bits)
-with an extra modulo operator that will turn (-1 % 3) into 2 instead of -1
-with an extra division operator that will turn (-1 / 2) into -1 instead of 0
-with more logical names for variable types, a int with at least 32 bits is now called "long", what's longer than a long, a "long long", what's longer than a "long long"? a "long long long" or "super long"? This might bet pathetic in the future, a naming sheme of the form "sint16", "uint64", "float128", ... would be a better alternative
-with UTF-8, UTF-16 and UTF-32 characters
-with a variable type that is encouraged to be used to represent 8-bit bytes and that is unsigned, many people have had confusions with chars being used for files and them being signed
-with "nan" and "inf" values being part of the language instead of part of the library
-with fixed point types
-...

With regard to Lode's point about iterator names, I believe an 'auto' data type is being discussed for C++ that would derive its type from the right-hand side of its initialisation:

vector<int> v;
for(auto i=v.begin();i<v.end();++i)

// etc

Would obviously require an initialisation, like a reference, but would save a lot of typing. Does anyone know if this idea is still being bandied around for the next standard?