1. "The Better String Library" *outperforms* standard C string library functions, std::string and MFC's CString class on a variety of the most common string manipulation tasks (often by massive margins). This has been confirmed by benchmarking on multiple CPUs, operating systems and compilers.

2. Avoiding unnecessary heap allocations is a misdirected effort. ("copy-on-write" is hard to make thread safe, and certainly cannot be done without impacting performance). It only helps in the most marginal cases while adding to the overhead of mainline code.

3. Inlining is usually only necessary and/or beneficial (performance-wise) for trivial accessor functions, while adding to the footprint of your code (since code paths are redundantly generated.) The best answer is just to use the standard "inline" declaration on methods where it is appropriate.

For the evidence I invite you to look at all the data here:

http://bstring.sf.net/features.html

You can even download the benchmarks for yourself from the examples.

But of course, this all ignores the far more important question about string manipulation in todays security and cost conscious programming environment:

A. How safe is the string manipulation with respect to buffer overflows?
B. How easy is the library to use and understand?
C. How portable is the library?
D. Is the library well documented?
E. Is the library well tested?
F. Is the library being actively maintained?

You can check the Better String library against any of these criteria, and you will find it hard to beat with *any* alternative.