Modern Network Unicode

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶

This Internet-Draft will expire on 3 March 2024.¶

Copyright Notice

Copyright (c) 2023 IETF Trust and the persons identified as the document authors. All rights reserved.¶

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶

Table of Contents

1. Introduction

(Insert embellished copy of abstract here.)¶

Complex specifications that use Unicode often come with detailed information on their Unicode usage; this level of detail generally is necessary to support some legacy applications. New, simple protocol specifications generally do not have such a legacy or need such details, but can instead simply use common practice, informed by decades of using Unicode. The present specification attempts to serve as a convenient reference for such protocol specifications, reducing their need for discussing Unicode to just pointing to the present specification and making a few simple choices.¶

There is no intention that henceforth all new protocols "must" use the present specification. It is offered as a standards-track specification simply so it can be normatively referenced from other standards-track specifications.¶

2. 1D Modern Network Unicode

One-dimensional Modern Network Unicode (1D MNU) is the form of Modern Network Unicode that can be used for one-dimensional text, i.e., text without a structure of separate text lines. It requires conformance to [STD63], as well as to the following four mandates:¶

1. Control codes (U+0000 to U+001F and U+007F to U+009F) MUST NOT be used. (Note that this also excludes line endings, so a 1D MNU text string cannot extend beyond a single line. See Section 3 below if line structure is needed.)¶
2. The characters U+2028 and U+2029 MUST NOT be used. (In case future Unicode versions add to the Unicode character categories Zl or Zp, any characters in these categories MUST NOT be used.)¶
3. Modern Network Unicode strongly RECOMMENDS that, except in unusual circumstances (see Appendix C), all text is transmitted in normalization form NFC. This mandate should not be realized by routinely normalizing all text, but by encouraging text sources to use NFC if applicable.¶
4. As per the Unicode specification, the code points U+FFFE and U+FFFF MUST NOT be used. Also, Byte Order Marks (leading U+FEFF characters) MUST NOT be used.¶

Note that several control codes have been in use historically in ASCII documents even within a single text line. For instance BS (U+0008) and CR (U+000D) have been used as a crude way to combine (by overtyping) characters; the present specification does not explicitly support this behavior anymore. HT (U+0009, "TAB") has been used for a form of compressing stretches of blank space; by keeping its actual definition open it has also been used to enable users of text to specify variable blank space ("indentation"). The present specification RECOMMENDS not using a variance for 1D MNU to enable using BS, HT, or CR; legacy usage of HT may be more appropriate in certain forms of 2D text.¶

3. 2D Modern Network Unicode

Two-dimensional Modern Network Unicode (2D MNU) enhances 1D MNU by structuring the text into lines. 2D MNU does this is by using the control code LF (U+000A) for a line terminator. The use of an unterminated last line is permitted, but not encouraged. (This is not a variance.)¶

Historically, the Internet NVT [RFC0854] and thus, much later, Network Unicode [RFC5198] have used the combination CR+LF as a line terminator, reflecting its heritage based on teletype functionality. The use of U+000D mostly introduces additional ways to create interoperability problems. [XML] goes to great lengths to reduce the harm the presence of CR characters does to interchange of XML documents. Pages 10 and 11 of [RFC0764] discuss how CR can actually only be used in the combinations CR+NUL and CR+LF. 2D MNU simply elides the CR.¶

In other words, 2D MNU adds the use of line endings, represented by a single LF character (which is then the only control character allowed). Variances are available for (1) allowing or even (2) requiring the use of CR before LF.¶

4. 3D?

Three-dimensional forms of text have been used, e.g., by using U+000C FF ("form feed") to add a page structure to the RFC format [RFC2223], or by using U+001E RS ("record separator") to separate several (2D) JSON texts in a JSON sequence [RFC7464].¶

As the need for this form of structure is now mostly being addressed by building data structures around text items, the present specification does not define a 3D MNU. If needed, variances such as the use of FF or RS can be described in the documents specifying the 3D format.¶

5. Variances

In addition to the basic 1D and 2D versions of MNU, this specification describes a number of variances that can be used in the forms such as "2D Modern Network Unicode with VVV", or "2D Modern Network Unicode with VVV, WWW, and ZZZ" for multiple variances used. Specifications that cannot directly use the basic MNU forms may be able to use MNU with one or more of these variances added.¶

An example for the usage of variances: Up to RFC 8649, RFCs were formatted in "2D Modern Network Unicode with CR-tolerant lines and FF characters", or exceptionally [RFC8265], "2D Modern Network Unicode with CR-tolerant lines and FF characters and BOM".¶

6. Using ABNF with Unicode

Internet STD 68, [STD68], defines Augmented BNF for Syntax Specifications: ABNF. Since the late 1970s, ABNF has often been used to formally describe the pieces of text that are meant to be used in an Internet protocol. ABNF was developed at a time when character coding grew more and more complicated, and even in its current form, discusses encoding of characters only briefly (Section 2.4 of [STD68]). This discussion offers no information about how this should be used today (it actually still refers to 16-bit Unicode!).¶

The best current practice of using ABNF for Unicode-based protocols is as follows: ABNF is used as a grammar for describing sequences of Unicode code points, valued from 0x0 to 0x10FFFF. The actual encoding (as UTF-8) is never seen on the ABNF level; see Section 9.4 of [RFC6020] for a recent example of this. Approaches such as representing the rules of UTF-8 encoding in ABNF (see Section 3.5 of [RFC5255] for an example) add complexity without benefit and are NOT RECOMMENDED.¶

ABNF features such as case-insensitivity in literal text strings essentially do not work for general Unicode; text string literals therefore (and by the definition in Section 2.3 of [STD68]) are limited to ASCII characters. That is often not actually a problem in text-based protocol definitions. Still, characters beyond ASCII need to be allowed in many productions. ABNF does not have access to Unicode character categories and thus will be limited in its expressiveness here. The core rules defines in Appendix B of [STD68] are limited to ASCII as well; new rules will therefore need to be defined in any protocol employing modern Unicode.¶

The present specification recommends defining ABNF rules as in these examples:¶

; 1D modern unicode character:
uchar = %x20-7E ; exclude C0
      / %xA0-2027 ; exclude DEL, C1
      / %x202A-D7FF ; exclude U+2028/2029
      / %xE000-FFFD ; exclude U+FFFE/FFFF
      / %x10000-10FFFD ; could exclude more non-characters

; 2D modern unicode with lines -- newline:
unl = %x0A
u2dchar = unl / uchar
; alternatively, CR-tolerant newline:
ucrnl = [%x0D] %x0A
u2dcrchar = ucrnl / uchar

; if really needed, HT-tolerant 1D unicode character:
utchar = %x09 / uchar

7. IANA considerations

This specification places no requirements on IANA.¶

8. Security considerations

The security considerations of [RFC5198] apply.¶

A variance "with NUL characters" would create specific security considerations as discussed in the security considerations of [RFC5198] and should therefore only be used in circumstances that absolutely do require it.¶

Appendix A. Terminology

In order for the main body of this specification to stay readable for its target audience, we banish what could be considered excessive detail into appendices. Expert readers who know a lot more about Coded Character Sets and Unicode than the target audience is likely interested in are requested to hold back the urge to request re-introduction of excessive detail into the main body. For instance, this specification has a local definition of "Unicode character" that is useful for protocol designers, but is not actually defined by the Unicode consortium; the details are here.¶

Some definitions below reference definitions given in the Unicode standard and are simplified from those, which see if the full detail is needed.¶

Grapheme:

The smallest functional unit of a writing system. For computer representation, graphemes are represented by characters, sometimes a single character, sometimes decomposed into multiple characters. In visible presentation, graphemes are made visible as glyphs; for computer use, glyphs are collected into typefaces (fonts). The term grapheme is used to abstract from the specific glyph used — a specific glyph from a typeface may be considered one of several allographs for a grapheme.¶

Character:

Only defined as "abstract character" in [UNICODE]: A unit of information used for the organization, control, or representation of textual data (definition D7). Most characters are used to stand for graphemes or to build graphemes out of several characters. Characters are usually used in ordered sequences, which are referred to as "character strings".¶

Character set, Coded Character Set:

A mapping from code points to characters. Note that the shorter form term "character set" is easily misunderstood for a set of characters, which are called "character repertoire"¶

Character repertoire:

A set of characters, in the sense of what is available as characters to choose from.¶

Code point:

The (usually numeric) value used in a coded character set to refer to a character. Note that code points may undergo some encoding before interchange, see Unicode transformation format below. Code points can also be allocated to refer to internal constructs of a Unicode transformation format instead of referring to characters.¶

Unicode uses unsigned numbers between 0 and 1114111 (0x10FFFF) as code points.¶

Unicode transformation format (UTF):

Character (and character strings) usually are interchanged as a sequence of bytes. A Unicode Transformation Format or Unicode Encoding Scheme specifies a byte serialization for a Unicode encoding form. (See definition D94 in Section 3.10, Unicode Encoding Schemes.)¶

Code unit:

A bit combination used for encoding text for processing or interchange. The Unicode Standard uses 8-bit code units in the UTF-8 encoding form, 16-bit code units in the UTF-16 encoding form, and 32-bit code units in the UTF-32 encoding form. (See Definition D77 in Section 3.9, Unicode Encoding Forms.)¶

Unicode scalar value:

Unicode code points except high-surrogate and low-surrogate code points (see Appendix B). In other words, the ranges of integers 0 to 0xD7FF and 0xE000 to 0x10FFFF16 inclusive. (See Definition D76 in Section 3.9, Unicode Encoding Forms.)¶

Unicode Character:

For the purposes of the present specification, we use "Unicode Character" as a synonym for "Unicode scalar value". Note that this term includes Unicode scalar values that are not Assigned Characters; for the purposes of a specification using the present document, it is rarely useful to make this distinction, as Unicode evolves and could assign characters not assigned.¶

Appendix B. History, Legacy

Some of the complexity of using Unicode is, unsurprisingly, rooted in its long history of development.¶

Unicode originally was introduced around 1988 as a 16-bit character standard. By the early 1990s, a specification had been published, and a number of environments eagerly picked up Unicode. Unicode characters were represented as 16-bit values (UCS-2), enabling a simple character model using these uniformly 16-bit values as the characters. Programming language environments such as those of Java, JavaScript and C# are now intimately entangled with this character model.¶

In the mid-1990s, it became clear that 16 bits would not suffice. Unicode underwent an extension to 31 bits, and then back to ~ 21 bits (0 to 0x10FFFF). For the 16-bit environments, this was realized by switching to UTF-16, a "Unicode transformation format" (UTF-16). This reassigned some of the 16-bit code points not for characters, but for 2 sets of 1024 "surrogates" that would be used in pairs to represent characters that do not fit into 16 bits, each supplying 10 bits to together add 2²⁰ code points to the 2¹⁶ already available (this leads to the surprising upper limit of 0x10FFFF).¶

In parallel, UTF-8 was created as an ASCII-compatible form of Unicode representation that would become the dominant form of text in the Web and other interchange environments.¶

The UCS-2 based character models of the legacy 16-bit platforms in many cases couldn't be updated for fully embracing UTF-16 right away. For instance, only much later did ECMAScript introduce the "u" (Unicode) flag for regular expressions to have them actually match "Unicode" characters. So, on these platforms, UTF-16 is handled in a UCS-2 character model, and sometimes orphaned surrogates leak out instead of Unicode characters as "code points" in interfaces that are not meant to impose these implementation limitations on the outside world.¶

UTF-8 is unaffected by this UTF-16 quirk and of course doesn't support encoding surrogates. (UTF-8 is careful to allow a single representation only for each Unicode character, and enabling the alternative use of surrogate pairs would violate that invariant, while isolated surrogates don't mean anything in Unicode.)¶

Newly designed IETF protocols typically do not have to consider these problems, but occasionally there are attempts to include isolated surrogate code points into what we call Unicode characters here.¶

Appendix C. Normalization

Please see Section 3 of [RFC5198] for a brief introduction to Unicode normalization and Normalization Form C (NFC). However, since that section was written, additional experience with normalization has led to the realization that the Unicode normalization rules do not always preserve certain details in certain writing systems.¶

Therefore, the implementation approach of routinely normalizing all text before interchange has fallen out of favor. Instead, implementations are encouraged to use text sources that already generally use NFC except where normalization would have been harmful. Where two text strings needs to be compared, it may be appropriate to employ normalization to both text strings for the purposes of the comparison only.¶

Additional complications can come from the fact that some implementations of applications may rely on operating system libraries over which they have little control. The need to maintain interoperability in such environments suggests that receivers of MNU should be prepared to receive unnormalized text and should not react to that in excessive ways; however, there also is no expectation for receivers to go out of their way doing so.¶

Appendix D. Relationship to RFC 5198

Of the mandates listed in Section 2, the third and fourth requirement are also posed by [RFC5198], while the first two remove further legacy compatibility considerations.¶

[RFC5198] contains some discussion and background material that the present document does not attempt to repeat; the interested reader may therefore want to consult it as an informative reference.¶

Mandates of [RFC5198] that are specific to a version of Unicode are not picked up in this specification, e.g., there is no check for unassigned code points. Some implementations may want to add such a check; however, in general, this can hinder further evolution as it may become hard to use new characters as long as not every component on the way has been upgraded. (see also Section 5.8.)¶

Acknowledgements

Klaus Hartke and Henk Birkholz drove the author out of his mind enough to make him finally write this up. James Manger, Tim Bray and Martin Thomson provided comments on an early version of this draft.¶

Author's Address