| Internet-Draft | CDDL grammar updates | October 2024 |
| Bormann | Expires 1 May 2025 | [Page] |
The Concise Data Definition Language (CDDL), as defined in RFC 8610 and RFC 9165, provides an easy and unambiguous way to express structures for protocol messages and data formats that are represented in CBOR or JSON.¶
The present document updates RFC 8610 by addressing errata and making other small fixes for the ABNF grammar defined for CDDL there.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://cbor-wg.github.io/update-8610-grammar/. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-ietf-cbor-update-8610-grammar/.¶
Discussion of this document takes place on the CBOR Working Group mailing list (mailto:cbor@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/cbor/. Subscribe at https://www.ietf.org/mailman/listinfo/cbor/.¶
Source for this draft and an issue tracker can be found at https://github.com/cbor-wg/update-8610-grammar.¶
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 1 May 2025.¶
Copyright (c) 2024 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.¶
The Concise Data Definition Language (CDDL), as defined in [RFC8610] and [RFC9165], provides an easy and unambiguous way to express structures for protocol messages and data formats that are represented in CBOR or JSON.¶
The present document updates [RFC8610] by addressing errata and making other small fixes for the ABNF grammar defined for CDDL there. The body of this document motivates and explains the updates; the updated collected ABNF syntax in Figure 11 in Appendix A replaces the collected ABNF syntax in Appendix B of [RFC8610].¶
A number of errata reports have been made around some details of text string and byte string literal syntax: [Err6527] and [Err6543]. These are being addressed in this section, updating details of the ABNF for these literal syntaxes. Also, [Err6526] needs to be applied (backslashes have been lost during RFC processing in some text explaining backslash escaping).¶
These changes are intended to mirror the way existing implementations have dealt with the errata. They also use the opportunity presented by the necessary cleanup of the grammar of string literals for a backward compatible addition to the syntax for hexadecimal escapes. The latter change is not automatically forward compatible (i.e., CDDL specifications that make use of this syntax do not necessarily work with existing implementations until these are updated, which this specification recommends).¶
The ABNF used in [RFC8610] for the content of text string literals is rather permissive:¶
; RFC 8610 ABNF: text = %x22 *SCHAR %x22 SCHAR = %x20-21 / %x23-5B / %x5D-7E / %x80-10FFFD / SESC SESC = "\" (%x20-7E / %x80-10FFFD)
This allows almost any non-C0 character to be escaped by a backslash,
but critically misses out on the \uXXXX and \uHHHH\uLLLL forms
that JSON allows to specify characters in hex (which should be
applying here according to Bullet 6 of Section 3.1 of [RFC8610]).
(Note that CDDL imports from JSON the unwieldy \uHHHH\uLLLL syntax,
which represents Unicode code points beyond U+FFFF by making them look
like UTF-16 surrogate pairs; CDDL text strings are not using UTF-16 or
surrogates.)¶
Both can be solved by updating the SESC rule.
This document uses the opportunity to add a popular form of directly specifying
characters in strings using hexadecimal escape sequences of the form
\u{hex}, where hex is the hexadecimal representation of the
Unicode scalar value.
The result is the new set of rules defining SESC in Figure 2:¶
; new rules collectively defining SESC:
SESC = "\" ( %x22 / "/" / "\" / ; \" \/ \\
%x62 / %x66 / %x6E / %x72 / %x74 / ; \b \f \n \r \t
(%x75 hexchar) ) ; \uXXXX
hexchar = "{" (1*"0" [ hexscalar ] / hexscalar) "}" /
non-surrogate / (high-surrogate "\" %x75 low-surrogate)
non-surrogate = ((DIGIT / "A"/"B"/"C" / "E"/"F") 3HEXDIG) /
("D" %x30-37 2HEXDIG )
high-surrogate = "D" ("8"/"9"/"A"/"B") 2HEXDIG
low-surrogate = "D" ("C"/"D"/"E"/"F") 2HEXDIG
hexscalar = "10" 4HEXDIG / HEXDIG1 4HEXDIG
/ non-surrogate / 1*3HEXDIG
HEXDIG1 = DIGIT1 / "A" / "B" / "C" / "D" / "E" / "F"
(Notes:
In ABNF, strings such as "A", "B" etc. are case-insensitive, as is
intended here.
The rules above could, instead of %x62, also have used %s"b" etc., but didn't, in order to
maximize ABNF tool compatibility.)¶
Now that SESC is more restrictively formulated, this also requires an update to the BCHAR rule used in the ABNF syntax for byte string literals:¶
; RFC 8610 ABNF: bytes = [bsqual] %x27 *BCHAR %x27 BCHAR = %x20-26 / %x28-5B / %x5D-10FFFD / SESC / CRLF bsqual = "h" / "b64"
With the SESC updated as above, \' is no longer allowed in BCHAR;
this now needs to be explicitly included; see below.¶
Updating BCHAR also provides an opportunity to address [Err6278], which points to an inconsistency in treating U+007F (DEL) between SCHAR and BCHAR. As U+007F is not printable, including it in a byte string literal is as confusing as for a text string literal, and it should therefore be excluded from BCHAR as it is from SCHAR. The same reasoning also applies to the C1 control characters, so the updated ABNF actually excludes the entire range from U+007F to U+009F. The same reasoning then also applies to text in comments (PCHAR). For completeness, all these should also explicitly exclude the code points that have been set aside for UTF-16's surrogates.¶
; new rules for SCHAR, BCHAR, and PCHAR: SCHAR = %x20-21 / %x23-5B / %x5D-7E / NONASCII / SESC BCHAR = %x20-26 / %x28-5B / %x5D-7E / NONASCII / SESC / "\'" / CRLF PCHAR = %x20-7E / NONASCII NONASCII = %xA0-D7FF / %xE000-10FFFD
(Note that, apart from addressing the inconsistencies, there is no attempt to further exclude non-printable characters from the ABNF; doing this properly would draw in complexity from the ongoing evolution of the Unicode standard that is not needed here.)¶
The above changes also cover [Err6543] (a proposal to split off qualified byte string literals from UTF-8 byte string literals) and [Err6526] (lost backslashes); see Appendix B for details.¶
The CDDL example in Figure 5 demonstrates various escaping
techniques now available for (byte and text) strings in CDDL.
Obviously in the literals for a and x, there is no need to escape
the second character, an o, as \u{6f}; this is just for demonstration.
Similarly, as shown in c and z there also is no need to escape the
🁳 or ⌘, but escaping them may be convenient in order to limit the character
repertoire of a CDDL file itself to ASCII [STD80].¶
start = [a, b, c, x, y, z]
; "🁳", DOMINO TILE VERTICAL-02-02, and
; "⌘", PLACE OF INTEREST SIGN, in a text string:
a = "D\u{6f}mino's \u{1F073} + \u{2318}" ; \u{}-escape 3 chars
b = "Domino's \uD83C\uDC73 + \u2318" ; escape JSON-like
c = "Domino's 🁳 + ⌘" ; unescaped
; in a byte string given as text, the ' needs to be escaped:
x = 'D\u{6f}mino\u{27}s \u{1F073} + \u{2318}' ; \u{}-escape 4 chars
y = 'Domino\'s \uD83C\uDC73 + \u2318' ; escape JSON-like
z = 'Domino\'s 🁳 + ⌘' ; escape ' only
In this example, the rules a to c and x to z all produce strings with
byte-wise identical content, where a to c are text strings, and x to z
are byte strings.
Figure 6 illustrates this by showing the output generated from
the start rule in Figure 5, using pretty-printed hexadecimal.¶
86 # array(6)
73 # text(19)
446f6d696e6f277320f09f81b3202b20e28c98 # "Domino's 🁳 + ⌘"
73 # text(19)
446f6d696e6f277320f09f81b3202b20e28c98 # "Domino's 🁳 + ⌘"
73 # text(19)
446f6d696e6f277320f09f81b3202b20e28c98 # "Domino's 🁳 + ⌘"
53 # bytes(19)
446f6d696e6f277320f09f81b3202b20e28c98 # "Domino's 🁳 + ⌘"
53 # bytes(19)
446f6d696e6f277320f09f81b3202b20e28c98 # "Domino's 🁳 + ⌘"
53 # bytes(19)
446f6d696e6f277320f09f81b3202b20e28c98 # "Domino's 🁳 + ⌘"
The two subsections in this section specify two small changes to the grammar that are intended to enable certain kinds of specifications. These changes are backward compatible, i.e., CDDL files that comply to [RFC8610] continue to match the updated grammar, but not necessarily forward compatible, i.e., CDDL specifications that make use of these changes cannot necessarily be processed by existing [RFC8610] implementations.¶
[RFC8610] requires a CDDL file to have at least one rule.¶
; RFC 8610 ABNF: cddl = S 1*(rule S)
This makes sense when the file has to stand alone, as a CDDL data model needs to have at least one rule to provide an entry point (start rule).¶
With CDDL modules [I-D.ietf-cbor-cddl-modules], CDDL files can also include directives, and these might be the source of all the rules that ultimately make up the module created by the file. Any other rule content in the file has to be available for directive processing, making the requirement for at least one rule cumbersome.¶
Therefore, the present update extends the grammar as in Figure 8 and turns the existence of at least one rule into a semantic constraint, to be fulfilled after processing of all directives.¶
; new top-level rule: cddl = S *(rule S)
The existing ABNF syntax for expressing tags in CDDL is:¶
; extracted from RFC 8610 ABNF:
type2 =/ "#" "6" ["." uint] "(" S type S ")"
This means tag numbers can only be given as literal numbers (uints).
Some specifications operate on ranges of tag numbers, e.g., [RFC9277]
has a range of tag numbers 1668546817 (0x63740101) to 1668612095
(0x6374FFFF) to tag specific content formats.
This can currently not be expressed in CDDL.
Similar considerations apply to simple values (#7.xx).¶
This update extends the syntax to:¶
; new rules collectively defining the tagged case:
type2 =/ "#" "6" ["." head-number] "(" S type S ")"
/ "#" "7" ["." head-number]
head-number = uint / ("<" type ">")
For #6, the head-number stands for the tag number.
For #7, the head-number stands for the simple value if it is in
the ranges 0..23 or 32..255 (as per Section 3.3 of RFC 8949 [STD94]
the simple values 24..31 are not used).
For 24..31, the head-number stands for the "additional
information", e.g., #7.25 or #7.<25> is a float16, etc.
(All ranges mentioned here are inclusive.)¶
So the above range can be expressed in a CDDL fragment such as:¶
ct-tag<content> = #6.<ct-tag-number>(content) ct-tag-number = 1668546817..1668612095 ; or use 0x63740101..0x6374FFFF¶
Notes:¶
This syntax reuses the angle bracket syntax for generics;
this reuse is innocuous as a generic parameter/argument only ever
occurs after a rule name (id), while it occurs after . here.
(Whether there is potential for human confusion can be debated; the
above example deliberately uses generics as well.)¶
The updated ABNF grammar makes it a bit more explicit that the
number given after the optional dot is special, not giving the CBOR
"additional information" for tags and simple values as it is with
other uses of # in CDDL.
(Adding this observation to Section 2.2.3 of [RFC8610] is the subject
of [Err6575]; it is correctly noted in Section 3.6 of [RFC8610].)
In hindsight, maybe a different character than the dot should have
been chosen for this special case, however changing the grammar
now would have been too disruptive.¶
The grammar fixes and updates in this document are not believed to create additional security considerations. The security considerations in Section 5 of [RFC8610] do apply, and specifically the potential for confusion is increased in an environment that uses a combination of CDDL tools some of which have been updated and some of which have not been, in particular based on Section 2.¶
Attackers may want to exploit such potential confusion by crafting CDDL models that are interpreted differently by different parts of a system. There will be a period of transition from the details that the [RFC8610] grammar handled in a less well-defined way, to the updated grammar defined in the present document. This transition might offer one, but not the only kind of opportunity for the kind of attack that relies on differences between implementations. Implementations that make use of CDDL models operationally already need to ascertain the provenance (and thus authenticity and integrity) and applicability of models they employ. At the time of writing, it is expected that the models will generally be processed by a software developer, within a software development environment. Developers are therefore advised to treat CDDL models with the same care as any other source code.¶
This document has no IANA actions.¶
This appendix is normative.¶
It provides the full ABNF from [RFC8610] with the updates applied in the present document.¶
cddl = S *(rule S)
rule = typename [genericparm] S assignt S type
/ groupname [genericparm] S assigng S grpent
typename = id
groupname = id
assignt = "=" / "/="
assigng = "=" / "//="
genericparm = "<" S id S *("," S id S ) ">"
genericarg = "<" S type1 S *("," S type1 S ) ">"
type = type1 *(S "/" S type1)
type1 = type2 [S (rangeop / ctlop) S type2]
; space may be needed before the operator if type2 ends in a name
type2 = value
/ typename [genericarg]
/ "(" S type S ")"
/ "{" S group S "}"
/ "[" S group S "]"
/ "~" S typename [genericarg]
/ "&" S "(" S group S ")"
/ "&" S groupname [genericarg]
/ "#" "6" ["." head-number] "(" S type S ")"
/ "#" "7" ["." head-number]
/ "#" DIGIT ["." uint] ; major/ai
/ "#" ; any
head-number = uint / ("<" type ">")
rangeop = "..." / ".."
ctlop = "." id
group = grpchoice *(S "//" S grpchoice)
grpchoice = *(grpent optcom)
grpent = [occur S] [memberkey S] type
/ [occur S] groupname [genericarg] ; preempted by above
/ [occur S] "(" S group S ")"
memberkey = type1 S ["^" S] "=>"
/ bareword S ":"
/ value S ":"
bareword = id
optcom = S ["," S]
occur = [uint] "*" [uint]
/ "+"
/ "?"
uint = DIGIT1 *DIGIT
/ "0x" 1*HEXDIG
/ "0b" 1*BINDIG
/ "0"
value = number
/ text
/ bytes
int = ["-"] uint
; This is a float if it has fraction or exponent; int otherwise
number = hexfloat / (int ["." fraction] ["e" exponent ])
hexfloat = ["-"] "0x" 1*HEXDIG ["." 1*HEXDIG] "p" exponent
fraction = 1*DIGIT
exponent = ["+"/"-"] 1*DIGIT
text = %x22 *SCHAR %x22
SCHAR = %x20-21 / %x23-5B / %x5D-7E / NONASCII / SESC
SESC = "\" ( %x22 / "/" / "\" / ; \" \/ \\
%x62 / %x66 / %x6E / %x72 / %x74 / ; \b \f \n \r \t
(%x75 hexchar) ) ; \uXXXX
hexchar = "{" (1*"0" [ hexscalar ] / hexscalar) "}" /
non-surrogate / (high-surrogate "\" %x75 low-surrogate)
non-surrogate = ((DIGIT / "A"/"B"/"C" / "E"/"F") 3HEXDIG) /
("D" %x30-37 2HEXDIG )
high-surrogate = "D" ("8"/"9"/"A"/"B") 2HEXDIG
low-surrogate = "D" ("C"/"D"/"E"/"F") 2HEXDIG
hexscalar = "10" 4HEXDIG / HEXDIG1 4HEXDIG
/ non-surrogate / 1*3HEXDIG
bytes = [bsqual] %x27 *BCHAR %x27
BCHAR = %x20-26 / %x28-5B / %x5D-7E / NONASCII / SESC / "\'" / CRLF
bsqual = "h" / "b64"
id = EALPHA *(*("-" / ".") (EALPHA / DIGIT))
ALPHA = %x41-5A / %x61-7A
EALPHA = ALPHA / "@" / "_" / "$"
DIGIT = %x30-39
DIGIT1 = %x31-39
HEXDIG = DIGIT / "A" / "B" / "C" / "D" / "E" / "F"
HEXDIG1 = DIGIT1 / "A" / "B" / "C" / "D" / "E" / "F"
BINDIG = %x30-31
S = *WS
WS = SP / NL
SP = %x20
NL = COMMENT / CRLF
COMMENT = ";" *PCHAR CRLF
PCHAR = %x20-7E / NONASCII
NONASCII = %xA0-D7FF / %xE000-10FFFD
CRLF = %x0A / %x0D.0A
This appendix is informative.¶
[Err6543] observes that the ABNF used in [RFC8610] for the content of byte string literals lumps together byte strings notated as text with byte strings notated in base16 (hex) or base64 (but see also updated BCHAR rule in Figure 4):¶
; RFC 8610 ABNF: bytes = [bsqual] %x27 *BCHAR %x27 BCHAR = %x20-26 / %x28-5B / %x5D-10FFFD / SESC / CRLF
Errata report 6543 proposes to handle the two cases in separate ABNF rules (where, with an updated SESC, BCHAR obviously needs to be updated as above):¶
; Err6543 proposal:
bytes = %x27 *BCHAR %x27
/ bsqual %x27 *QCHAR %x27
BCHAR = %x20-26 / %x28-5B / %x5D-10FFFD / SESC / CRLF
QCHAR = DIGIT / ALPHA / "+" / "/" / "-" / "_" / "=" / WS
This potentially causes a subtle change, which is hidden in the WS rule:¶
; RFC 8610 ABNF: WS = SP / NL SP = %x20 NL = COMMENT / CRLF COMMENT = ";" *PCHAR CRLF PCHAR = %x20-7E / %x80-10FFFD CRLF = %x0A / %x0D.0A
This allows any non-C0 character in a comment, so this fragment becomes possible:¶
foo = h' 43424F52 ; 'CBOR' 0A ; LF, but don't use CR! '¶
The current text is not unambiguously saying whether the three apostrophes
need to be escaped with a \ or not, as in:¶
foo = h' 43424F52 ; \'CBOR\' 0A ; LF, but don\'t use CR! '¶
... which would be supported by the existing ABNF in [RFC8610].¶
This document takes the simpler approach of leaving the processing of
the content of the byte string literal to a semantic step after
processing the syntax of the bytes/BCHAR rules, as updated by
Figure 2 and Figure 4 in Section 2.1 (updates prompted by the combination
of [Err6527] and [Err6278]).¶
The rules in Figure 14 (as updated by Figure 4) are therefore
applied to the result of this
processing where bsqual is given as h or b64.¶
Note that this approach also works well with the use of byte strings
in Section 3 of [RFC9165].
It does require some care when copy-pasting into CDDL models from ABNF
that contains single quotes (which may also hide as apostrophes
in comments); these need to be escaped or possibly replaced by %x27.¶
Finally, the approach taken lends support to extending bsqual in CDDL
similar to the way this is done for CBOR diagnostic notation in [I-D.ietf-cbor-edn-literals].
(Note that the processing of string literals now is quite similar between
CDDL and EDN, except that CDDL has ";"-based end-of-line comments, while EDN has
two comment syntaxes, in-line "/"-based and end-of-line "#"-based.)¶
Many thanks go to the submitters of the errata reports addressed in this document. In one of the ensuing discussions, Doug Ewell proposed to define an ABNF rule NONASCII, of which we have included the essence. Special thanks to the reviewers Marco Tiloca, Christian Amsüss (shepherd review and further guidance), Orie Steele (AD review and further guidance), and Éric Vyncke (detailed IESG review).¶