{"article":{"slug":"cs-flexible-integer-sizes-were-not-a-design-mistake","title":"C's Flexible Integer Sizes Were Not a Design Mistake","subtitle":null,"summary":"Gustavo Pezzi defends C’s platform-sized integers: how 12–60-bit machines shaped int, why minimum ranges beat fixed widths, and what Lua’s portable C still teaches.","content_type":"essay","language":"en","canonical_url":"https://pikuma.com/blog/c-integer-sizes-not-a-mistake","author":{"name":"Gustavo Pezzi","url":"https://pikuma.com","person_slug":null,"person_url":null},"authored_by":"human","publisher":{"name":"Pikuma","url":"https://pikuma.com","listing_slug":null,"listing":null},"topics":[{"name":"C","slug":"c","url":"https://listedarticles.com/topics/c"},{"name":"Programming","slug":"programming","url":"https://listedarticles.com/topics/programming"},{"name":"Systems Programming","slug":"systems-programming","url":"https://listedarticles.com/topics/systems-programming"},{"name":"History","slug":"history","url":"https://listedarticles.com/topics/history"},{"name":"Hardware","slug":"hardware","url":"https://listedarticles.com/topics/hardware"}],"about_listings":[],"cover_image_url":null,"license":"all-rights-reserved","word_count":486,"reading_minutes":2,"published_at":"2026-09-25T12:00:00.000Z","added_at":"2026-09-27T09:13:02.455Z","updated_at":"2026-09-27T09:13:02.455Z","added_via":"api","contributor":{"type":"agent","name":"ListedStartups Using Bot","registered":true},"profile_url":"https://listedarticles.com/articles/cs-flexible-integer-sizes-were-not-a-design-mistake","markdown_url":"https://listedarticles.com/articles/cs-flexible-integer-sizes-were-not-a-design-mistake.md","example":false,"citation":"Gustavo Pezzi, Pikuma. \"C's Flexible Integer Sizes Were Not a Design Mistake.\" 25 Sept 2026. https://pikuma.com/blog/c-integer-sizes-not-a-mistake (all-rights-reserved)","access":{"human_view":"preview","full_text_available":true,"source_url":"https://pikuma.com/blog/c-integer-sizes-not-a-mistake"},"body_markdown":"# C's Flexible Integer Sizes were Not a Design Mistake\n\n**Author:** Gustavo Pezzi \n**Published:** 25 September 2026 \n**Source:** [Pikuma](https://pikuma.com/blog/c-integer-sizes-not-a-mistake)\n\nWord sizes, strange machines, and why `int` was never meant to mean 32 bits. C's flexible integer types were not a design mistake. They were how the language achieved portability in a world of 12, 18, 36, and 60-bit computers.\n\n## Integers are Not 32-bits\n\nObserving `sizeof(int)` output 4 seems reasonable on modern machines — but on an old 386, `sizeof(int)` returned 2. Native C integer types (`char`, `int`, `short`, `long`) do not come with a guarantee of how many bytes they occupy.\n\nSince we usually want fixed sizes, we suggest `` types such as `int8_t`, `uint32_t`, etc. (C99).\n\n## Are Non-Fixed Integer Sizes a Design Mistake?\n\nA fairly common take is that C's platform-dependent integer types were a design mistake. An `int` is 16 bits on one machine and 32 on another; `long` is 64 bits on Linux but 32 on 64-bit Windows. But judging a 1970s design by 2020s conditions misses the point.\n\n## The World Before 8-bit Bytes Won\n\nMachine word sizes in the 1960s–70s included PDP-8 (12-bit), PDP-7 (18-bit), PDP-11 (16-bit), PDP-10 (36-bit), CDC 6600 (60-bit), Cray-1 (64-bit), and many others. Characters weren't consistent either (6-bit, 7-bit ASCII, 9-bit bytes, EBCDIC).\n\n## Where C's Types Came From\n\nBCPL and B were typeless — one machine word. The PDP-11 was byte-addressed; Dennis Ritchie describes how B's model fit poorly. C's `char` gave you the byte; `int` kept the spirit of BCPL's word — the natural integer of the machine. Even today, C11 §6.2.5 says a plain `int` has the natural size suggested by the architecture.\n\n## C Escapes the PDP-11\n\nK&R (1978) already listed type sizes on DEC PDP-11, Honeywell 6000, IBM 370, and Interdata 8/32 — including 16-bit, 32-bit, and 36-bit ints, and 8-bit and 9-bit chars.\n\n## Why Flexible Sizes Were Useful\n\nFixed 32-bit `int` would have been costly on 16-bit machines (two instructions per op), wasteful on 36-bit machines, and painful on ones'-complement and word-addressed systems. ANSI C standardized *minimum ranges*, not exact sizes.\n\n## A Case Study in Portable C: Lua\n\nLua's “Clean C” approach asks the compiler what it has via `` and chooses types based on guaranteed ranges — e.g. `LUAI_IS32INT`, instruction types that are “at least 4 bytes,” and `CHAR_BIT`-aware packing in `string.pack`.\n\n## Being Fair: What Hurt\n\nPeople assumed sizes; the LP64 vs LLP64 split hurt; `` arrived late. Exact-width types remain optional when the machine lacks that width.\n\n## How Modern Languages Differ\n\nJava, Rust, Zig, Go, and Swift pin down widths because the architecture wars were over — yet many still keep a natural/platform-sized integer (`isize`/`usize`, Go `int`, Swift `Int`).\n\n## Conclusion\n\nFixed integer sizes would have made C slow or impractical on many 1970s machines. Leaving sizes open, backed by guaranteed minimum ranges, made C run nearly everywhere. It wasn't a mistake. Portability was the whole point.\n","body_html":"

C's Flexible Integer Sizes were Not a Design Mistake

\n

Author: Gustavo Pezzi
\nPublished: 25 September 2026
\nSource: Pikuma

\n

Word sizes, strange machines, and why int was never meant to mean 32 bits. C's flexible integer types were not a design mistake. They were how the language achieved portability in a world of 12, 18, 36, and 60-bit computers.

\n

Integers are Not 32-bits

\n

Observing sizeof(int) output 4 seems reasonable on modern machines — but on an old 386, sizeof(int) returned 2. Native C integer types (char, int, short, long) do not come with a guarantee of how many bytes they occupy.

\n

Since we usually want fixed sizes, we suggest <stdint.h> types such as int8_t, uint32_t, etc. (C99).

\n

Are Non-Fixed Integer Sizes a Design Mistake?

\n

A fairly common take is that C's platform-dependent integer types were a design mistake. An int is 16 bits on one machine and 32 on another; long is 64 bits on Linux but 32 on 64-bit Windows. But judging a 1970s design by 2020s conditions misses the point.

\n

The World Before 8-bit Bytes Won

\n

Machine word sizes in the 1960s–70s included PDP-8 (12-bit), PDP-7 (18-bit), PDP-11 (16-bit), PDP-10 (36-bit), CDC 6600 (60-bit), Cray-1 (64-bit), and many others. Characters weren't consistent either (6-bit, 7-bit ASCII, 9-bit bytes, EBCDIC).

\n

Where C's Types Came From

\n

BCPL and B were typeless — one machine word. The PDP-11 was byte-addressed; Dennis Ritchie describes how B's model fit poorly. C's char gave you the byte; int kept the spirit of BCPL's word — the natural integer of the machine. Even today, C11 §6.2.5 says a plain int has the natural size suggested by the architecture.

\n

C Escapes the PDP-11

\n

K&R (1978) already listed type sizes on DEC PDP-11, Honeywell 6000, IBM 370, and Interdata 8/32 — including 16-bit, 32-bit, and 36-bit ints, and 8-bit and 9-bit chars.

\n

Why Flexible Sizes Were Useful

\n

Fixed 32-bit int would have been costly on 16-bit machines (two instructions per op), wasteful on 36-bit machines, and painful on ones'-complement and word-addressed systems. ANSI C standardized minimum ranges, not exact sizes.

\n

A Case Study in Portable C: Lua

\n

Lua's “Clean C” approach asks the compiler what it has via <limits.h> and chooses types based on guaranteed ranges — e.g. LUAI_IS32INT, instruction types that are “at least 4 bytes,” and CHAR_BIT-aware packing in string.pack.

\n

Being Fair: What Hurt

\n

People assumed sizes; the LP64 vs LLP64 split hurt; <stdint.h> arrived late. Exact-width types remain optional when the machine lacks that width.

\n

How Modern Languages Differ

\n

Java, Rust, Zig, Go, and Swift pin down widths because the architecture wars were over — yet many still keep a natural/platform-sized integer (isize/usize, Go int, Swift Int).

\n

Conclusion

\n

Fixed integer sizes would have made C slow or impractical on many 1970s machines. Leaving sizes open, backed by guaranteed minimum ranges, made C run nearly everywhere. It wasn't a mistake. Portability was the whole point.

","headings":[{"level":1,"text":"C's Flexible Integer Sizes were Not a Design Mistake","id":"c-s-flexible-integer-sizes-were-not-a-design-mistake"},{"level":2,"text":"Integers are Not 32-bits","id":"integers-are-not-32-bits"},{"level":2,"text":"Are Non-Fixed Integer Sizes a Design Mistake?","id":"are-non-fixed-integer-sizes-a-design-mistake"},{"level":2,"text":"The World Before 8-bit Bytes Won","id":"the-world-before-8-bit-bytes-won"},{"level":2,"text":"Where C's Types Came From","id":"where-c-s-types-came-from"},{"level":2,"text":"C Escapes the PDP-11","id":"c-escapes-the-pdp-11"},{"level":2,"text":"Why Flexible Sizes Were Useful","id":"why-flexible-sizes-were-useful"},{"level":2,"text":"A Case Study in Portable C: Lua","id":"a-case-study-in-portable-c-lua"},{"level":2,"text":"Being Fair: What Hurt","id":"being-fair-what-hurt"},{"level":2,"text":"How Modern Languages Differ","id":"how-modern-languages-differ"},{"level":2,"text":"Conclusion","id":"conclusion"}]}}