Why spaces for indentation?

[mikeball]

A general question about why decision was made to use spaces for indentation? Tabs would seem to save tokens, should negate need for configurable number of spaces by allowing differential visual indenting based on user editor preferences, and should avoid indentation copy/paste conflicts between differently configured indentations. Section 12 of SPEC.md:

• Encoders MUST use a consistent number of spaces per level (default 2; configurable).
• Tabs MUST NOT be used for indentation.

• https://github.com/toon-format/spec/blob/main/SPEC.md#12-indentation-and-whitespace

Did tabs cause an issue with other parts of the encoding? Thanks.

[johannschopplich]

Good question! We forbid tabs in indentation mainly for determinism and simplicity, not because it was impossible to support them.

• Token‑wise, indentation is a very small slice of the context, and GPT‑style tokenizers already have special tokens for "newline + spaces" Replacing " " with "\t" cuts characters, but in practice barely moves token counts compared to the big savings TOON already gets from not repeating keys and dropping JSON punctuation.
• Structurally, reserving tab only for "cell delimiter or string content" and reserving leading whitespace only for "spaces = indentation" makes both the spec and implementations simpler. You never have to ask "is this tab a delimiter or is it part of the indent?" and you avoid YAML‑style mixed tab/space headaches.
• From a tooling perspective, spaces render consistently across GitHub, editors, and chat UIs. Tabs have editor‑dependent widths and are more likely to be auto‑normalized, which is not great for a data format we want to look identical everywhere.
  That's why the canonical encoding is "spaces only" for indentation, and tabs are reserved for delimiters and quoted text.

That said, the spec does allow decoders in non‑strict mode to accept tabs in indentation with implementation‑defined rules (§12). So a library can choose to be lenient with tab‑indented input, while still emitting spaces for canonical output.

Hope that answers all of your questions!