bug#11187: fmt: doesn't understand multibyte characters (UTF-8)

"Dmitry E. Oboukhov" <[email protected]> Wed, 4 Mar 2026 22:58:53 +0300

Newsgroups	gmane.comp.gnu.core-utils.bugs
Message-ID	<aaiO_b2i-VolxlZP@nbw>

> Nowadays of course there's no real reason to use for fmt, as text
> editors now do that stuff for you.

That is where you are mistaken. You are thinking of a human sitting in
front of a modern IDE. But in 2026, the most active "users" of classic
CLI utilities like grep, sed, awk, and fmt are **AIs**.

The reason AI "advertises" fmt is because it expects a standard system
utility to perform its documented function. In fact, that is exactly
how I discovered this piece of "legacy" junk. I tasked an AI with a
formatting job, saw a mangled, broken output, and realized the AI had
naively trusted 'fmt' to handle 21st-century text.

When an AI agent processes data via a shell, it doesn't have an IDE.
It relies on the core building blocks of the system. If those blocks
are broken or stuck in the 1970s, the AI's output becomes a mess for
the human recipient.

If 'fmt' is a "useless program" that only exists because it's a
"hassle to remove," then why is it still being shipped in 2025/2026 as
a functional part of GNU Coreutils? Keeping a broken tool in the box
just because "it's a hassle to take it out" is exactly the kind of
technical debt that makes modern systems feel like they are built on
quicksand.

If it's a corpse, bury it. If it's a tool, fix its 14-year-old Unicode
blindness.

Best regards,
A human who has to clean up after 'fmt' and its 7-bit dreams.

signature.asc (application/pgp-signature, 833 B)