2.2.2 Record Size Encoding

For the variable-sized records that are used by this protocol, the record needs to contain the size, in octets, of the content. An implementation SHOULD support record sizes as large as 0xffffffff octets (encoded size requires five octets).<2>

As represented in the following figure, the encoding algorithm takes the size of the record payload as input in little-endian format and generates a stream of octets. The octets MUST be sent in the order in which they are generated.

Figure 2: The encoding algorithm

The following table lists the encoded sizes for the range of values of Size, which is computed as previously explained. The network ordering of octets is top-down. For example, if the size is in the range 0x80-0x3FFF, the network ordering of encoded size octets is (Size & 0x7F) | 0x80 followed by Size >> 0x07.

Integer value (size)	Encoding
0x00-0x7F	Size
0x80-0x3FFF	(Size & 0x7F)| 0x80 Size >> 0x07
0x4000-0x1FFFFF	(Size & 0x7F)| 0x80 ((Size >> 0x07) & 0x7F)| 0x80 Size >> 0x0E
0x200000-0x0FFFFFFF	(Size & 0x7F)| 0x80 ((Size >> 0x07) & 0x7F)| 0x80 ((Size >> 0x0E) & 0x7F)| 0x80 Size >> 0x15
0x10000000-0x0FFFFFFFF	(Size & 0x7F)| 0x80 ((Size >> 0x07) & 0x7F)| 0x80 ((Size >> 0x0E) & 0x7F)| 0x80 ((Size >> 0x15) & 0x7F)| 0x80 Size >> 0x1C

In the preceding table, "&" refers to a bitwise "and" operation, "|" refers to a bitwise "or" operation, and ">>" refers to a right-shift operation.