The buffering scheme I've come up with is simplicity itself.  Rather than allocating increasingly large buffers only to move the previous buffer into the new buffer I've taken a different approach.   My buffer is a string array.  When one string fills up the class simply moves on to the next string in the array.   While this seems like a minor change it has two very profound effects on string concatenation.  First, this approach speeds concatenation up dramatically.  This is because buffers are never moved around which obviously takes time.  Second, much larger strings are possible because unlike conventional buffering schemes paged buffering does not require one huge and continuous chunk of memory.