Null-terminated strings were produced by the .ASCIZ directive of the PDP-11assembly languages and the ASCIZ directive of the MACRO-10 macro assembly language for the PDP-10. These predate the development of the C programming language, but other forms of strings were often used. At the time C was developed, memory was extremely limited, so using only one byte of overhead to store the length of a string was attractive. The only popular alternative at that time, usually called a "Pascal string", used a leading byte to store the length of the string. This allows the string to contain NUL and made finding the length need only one memory access , but limited string length to 255 characters. C designer Dennis Ritchie chose to follow the convention of NUL-termination, already established in BCPL, to avoid the limitation on the length of a string and because maintaining the count seemed, in his experience, less convenient than using a terminator. This had some influence on CPU instruction set design. Some CPUs in the 1970s and 1980s, such as the Zilog Z80 and the DECVAX, had dedicated instructions for handling length-prefixed strings. However, as the NUL-terminated string gained traction, CPU designers began to take it into account, as seen for example in IBM's decision to add the "Logical String Assist" instructions to the ES/9000 520 in 1992. FreeBSD developer Poul-Henning Kamp, writing in ACM Queue, would later refer to the victory of null-terminated strings over a 2-byte length as "the most expensive one-byte mistake" ever.
Limitations
While simple to implement, this representation has been prone to errors and performance problems. The NUL termination has historically created security problems. A NUL byte inserted into the middle of a string will truncate it unexpectedly. A common bug was to not allocate the additional space for the NUL, so it was written over adjacent memory. Another was to not write the NUL at all, which was often not detected during testing because a NUL was already there by chance from previous use of the same block of memory. Due to the expense of finding the length, many programs did not bother before copying a string to a fixed-size buffer, causing a buffer overflow if it was too long. The inability to store a NUL requires that string data and binary data be kept distinct and handled by different functions. This can lead to code redundancy and errors when the wrong function is used. The speed problems with finding the length can usually be mitigated by combining it with another operation that is O anyway, such as in strlcpy. However, this does not always result in an intuitive API.
Character encodings
Null-terminated strings require that the encoding does not use a zero byte anywhere, therefore it is not possible to store every possible ASCII or UTF-8 string. However, it is common to store the subset of ASCII or UTF-8 – every character except the NUL character – in null-terminated strings. Some systems use "modified UTF-8" which encodes the NUL character as two non-zero bytes and thus allow all possible strings to be stored. This is not allowed by the UTF-8 standard, because it is an overlong encoding, and it is seen as a security risk. Some other byte may be used as end of string instead, like 0xFE or 0xFF, which are not used in UTF-8. UTF-16 uses 2-byte integers and as either byte may be zero, cannot be stored in a null-terminated byte string. However, some languages implement a string of 16-bit UTF-16 characters, terminated by a 16-bit NUL character..
Improvements
Many attempts to make C string handling less error prone have been made. One strategy is to add safer functions such as strdup and strlcpy, whilst deprecating the use of unsafe functions such as gets. Another is to add an object-oriented wrapper around C strings so that only safe calls can be done. However, it is possible to call the unsafe functions anyway. Most modern libraries replace C strings with a structure containing a 32-bit or larger length value, and often add another pointer, a reference count, and even a NUL to speed up conversion back to a C string. Memory is far larger now, such that if the addition of 3 bytes to each string is a real problem the software will have to be dealing with so many small strings that some other storage method will save even more memory. Examples include the C++Standard Template Librarystd::string, the QtQString, the MFCCString, and the C-based implementation CFString from Core Foundation as well as its Objective-C sibling NSString from Foundation, both by Apple. More complex structures may also be used to store strings such as the rope.
