Using the original string - whether it's a literal in the source, part of a memory-mapped file, or even an allocated string "owned" by another part of your program - has the advantage of saving memory, and possibly eliminating ugly error conditions you'd otherwise have to handle if you performed an allocation (which could fail). The disadvantage, of course, is that you have to keep track of the fact that this string is not "owned" by the code currently using it, and thus that it cannot be modified/freed. Sometimes this means you need a flag in a structure to indicate whether a string it uses was allocated for the structure or not. With smaller programs, it might just mean you have to manually follow the logic of string ownership through several functions and make sure it's correct.
By the way, if the string is going to be used by a structure, one nice way to get around having to keep a flag marking whether it was allocated for the structure or not is to allocate space for the structure and the string (if needed) with a single call to malloc
. Then, freeing the structure always just works, regardless of whether the string was allocated for the structure or assigned from a string literal or other source.