Why is the buffer size in inet_ntoa 18?

Question

I looked at the implementations of inet_ntoa like this and this,

I am wondering why they both allocate a buffer of 18 characters.

If I take the maximum length IPv4 address string: 255.255.255.255 the size that I need is: 3 for each octet, 3 for dots and 1 for the null terminator. 3*4+3+1 = 16.

So why do we need those 2 extra characters?

The inet_ntoa implementation from the first link:

static __thread char buffer[18];

char *
inet_ntoa (struct in_addr in)
{
  unsigned char *bytes = (unsigned char *) ∈
  __snprintf (buffer, sizeof (buffer), "%d.%d.%d.%d",
              bytes[0], bytes[1], bytes[2], bytes[3]);
  return buffer;
}

Answer 1

If I take the maximum length IPv4 address string: 255.255.255.255 the size that I need is: 3 for each octet, 3 for dots and 1 for the null terminator. 3*4+3+1 = 16.

So why do we need those 2 extra characters?

Your computation is correct. Only sixteen bytes are needed for storing the dotted-decimal address string produced by inet_ntoa(), including its terminator. The relevant documentation and specifications specify the current format at least as far back as POSIX.1 2004, and as far as I am aware, no implementation has ever been released that produced any other format, so we can only speculate about why some implementations provide extra space. Possibilities include, but are not necessarily limited to

• a typo or miscalculation;
• the buffer was (maybe still is) used for more than one thing, and other uses require a couple more bytes;
• an algorithm used by some implementation to format the result benefited from having a couple of extra bytes to work with;
• extra bytes were provided for slop space, to mitigate the effects of hypothetical bugs.

That the same extra bytes are observed today in many implementations may support the multi-use buffer alternative, but that observation is also consistent with any explanation for those bytes appearing in some early implementation, maybe BSD, and being propagated from there to many subsequent ones. I'm inclined to favor the latter explanation.