Tracking String Lengths

Abstract

Programs that provide Internet services such as email, web browsing, and remote login communicate over the network by sending streams of bytes. Most of these exchanges are interpreted as strings (sequences of characters) that denote commands or requests. Parsing these commands is a particular challenge in servers written in C since the received byte stream is retrieved in chunks, where each chunk has an explicit size. Using chunks of memory with explicit size stands in contrast to the convention of standard string functions in C, which expect the length of a string to be determined by a nul character (a zero byte) at the end of a string. Mixing these two conventions can lead to subtle bugs in the program that do not show up until, for example, a malicious attacker sends a request string that contains a nul character. Another example is the program presented in the introduction (Fig. 1.2), which is incorrect on many platforms yet probably works seamlessly on most inputs since characters larger than 127 are rarely encountered in text files.