This is an old revision of the document!
Author names: Valid Unicode is required, and for non-ASCII names, an ASCII-only identifier is required.
Bibliographic text: The reference must point to something that has been translated to English; there must be at least an English language title and author. This applies to both normative and informative references.
Keywords: US-ASCII only
Body: The mention of non-ASCII characters requires identifiers, encourage characters, allow unicode names. (Note: use versus mention distinction)
Tables: Tables follow the same rules for identifiers and characters as the body. If it is sensible (i.e., more understandable for a reader) for a given document to have two tables, one including the identifiers and characters, one with just the characters, that will be allowed on a case by case basis.
U+ except within a code component where you must follow the rules of the programming language in which you are writing the code
Normalization forms: If the normalization matters to the content, the authors must submit in a normalization-resistant form. Do not expect normalization forms to be preserved.
All documents should identify themselves as being UTF-8. Both the canonical XML format and the non-canonical HTML format must contain metadata that specifies that the encoding is UTF-8. The non-canonical text-only format must begin with a UTF-8 BOM.
An implementer must be able to implement the specification without any confusion or ambiguity introduced by the use of UTF-8 rather than ASCII.
People must be able to reference (cite) the RFC from elsewhere in a standard way, including from documents that only support ASCII.
The RFC must be able to reference (cite) other documents in an unambiguous way.
Cross-references (including references to other documents) must be unambiguous even from a printed document.
Tools must be able to index the RFC in various ways, so searching for keywords, author names, and so on can work.