Agree on types and four-byte units

The data description supplies types; encoded values are not a dump of native memory.

• XDR means External Data Representation. It is a language for describing data and an encoding for its values, used by protocols such as ONC RPC and NFS. It is not itself a transport or an authentication protocol. A shared description, often called a schema, tells an encoder and decoder what types and order to expect (§1).
• XDR uses implicit typing: the selected example sends encoded values without inserting a type name before every value. The surrounding application already knows the description. An application can explicitly carry a description or introduce type tags using a union, but those are additional choices, not an automatic XDR header (§§1,5 discussion6).
• Every ordinary encoded item occupies a multiple of four bytes; void occupies zero bytes. Bytes are read in order. Data needing a non-multiple of four is followed by zero to three residual ZERO bytes, enough to reach the next four-byte boundary (§§3,4.16). This is value encoding, not a promise about native memory addresses.
• Integer encodings put the most significant byte first. The illustrative unsigned integer0x01020304 is sent as 01 02 03 04 regardless of native host byte order (§§4.1–4.2,5). The byte-order discussion is distinct from the physical bit order used by an Ethernet medium (§1).
• RFC 4506 is the May 2006 Internet Standard, STD 67. It obsoletes RFC 1832 but explicitly makes no technical encoding changes: §2 says it adds IANA/security discussion and distinguishes normative/informative references. Do not invent an XDR version change or change older encoded values.