C++ proposal: There are exactly 8 bits in a byte

Original Article ↗ Hacker News Discussion ↗

Motivation and Practical Reality

  • Many commenters note that virtually all modern systems already use 8‑bit bytes, and large amounts of C/C++ code implicitly assume this.
  • Several argue that formalizing CHAR_BIT == 8 simply matches de‑facto reality and improves developer expectations and portability in practice.
  • Others point out they only recently learned that byte was not guaranteed to be 8 bits, and see this as a surprising and confusing corner of the standard.

Legacy and Exotic Architectures

  • Multiple examples of non‑8‑bit “bytes” are discussed: 6‑, 7‑, 9‑, 10‑, 12‑, 16‑, 24‑, 32‑, and 36‑bit addressed units (UNIVAC, PDP‑10, various DSPs, historical mainframes, retro consoles).
  • Some of these systems still exist as emulated mainframes or niche DSPs, but often don’t track modern C++ standards or run mostly non‑C++ code.
  • A few participants enjoy hardware diversity and are sad to see standards drop support for such machines, even if they’re niche.

DSPs and Word‑Addressed Systems

  • DSP chips with 16‑ or 32‑bit addressable units are repeatedly cited as the only current plausible targets that conflict with CHAR_BIT == 8.
  • Opinions diverge: some say these platforms are specialized enough that non‑standard or frozen C/C++ dialects are acceptable; others insist they remain “real” systems that will lose conformance.

Portability vs. Simplification

  • Supporters frame the change as analogous to mandating two’s complement integers: reflect how “every real computer works now,” and reduce undefined edge cases.
  • Critics worry about excluding legitimate architectures and question the concrete benefit, since CHAR_BIT would still exist and much code already assumes 8.
  • Some see this as part of C++’s tension between maximal portability and acknowledging the practical hardware baseline.

Terminology and Related Debates

  • Several comments clarify: in C/C++ a “byte” is the size of char; “octet” is the unambiguously 8‑bit term, especially in networking.
  • There is side discussion about the fuzziness and obsolescence of “word” size, and about whether higher‑level code should think in bits, bytes, or fixed‑width integer types.