Additional Control Operators for CDDL

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 25 April 2022.¶

Copyright Notice

Copyright (c) 2021 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 Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.¶

Table of Contents

1. Introduction

The Concise Data Definition Language (CDDL), standardized in [RFC8610], provides "control operators" as its main language extension point (Section 3.8 of [RFC8610]).¶

The present document defines a number of control operators that were not yet ready at the time RFC 8610 was completed:¶

2. Computed Literals

CDDL as defined in [RFC8610] does not have any mechanisms to compute literals. To cover a large part of the use cases, this specification adds three control operators: .plus for numeric addition, .cat for string concatenation, and .det for string concatenation with dedenting of both sides (target and controller).¶

For these operators, as with all control operators, targets and controllers are types. The resulting type is therefore formally a function of the elements of the cross-product of the two types. Not all tools may be able to work with non-unique targets or controllers.¶

interval<BASE> = (
  BASE => int             ; lower bound
  (BASE .plus 1) => int   ; upper bound
  ? (BASE .plus 2) => int ; tolerance
)

X = 0
Y = 3
rect = {
  interval<X>
  interval<Y>
}

a = "foo" .cat '
  bar
  baz
'
; on a system where the newline is \n, is the same string as:
b = "foo\n  bar\n  baz\n"

oid = bytes .abnfb ("oid" .det cbor-tags-oid)
roid = bytes .abnfb ("roid" .det cbor-tags-oid)

cbor-tags-oid = '
  oid = 1*arc
  roid = *arc
  arc = [nlsb] %x00-7f
  nlsb = %x81-ff *%x80-ff
'

oid = bytes .abnfb 'oid
oid = 1*arc
roid = *arc
arc = [nlsb] %x00-7f
nlsb = %x81-ff *%x80-ff
'

3. Embedded ABNF

Many IETF protocols define allowable values for their text strings in ABNF [RFC5234] [RFC7405]. It is often desirable to define a text string type in CDDL by employing existing ABNF embedded into the CDDL specification. Without specific ABNF support in CDDL, that ABNF would usually need to be translated into a regular expression (if that is even possible).¶

ABNF is added to CDDL in the same way that regular expressions were added: by defining a .abnf control operator. The target is usually text or some restriction on it, the controller is the text of an ABNF specification.¶

There are several small issues, with solutions given here:¶

• ABNF can be used to define byte sequences as well as UTF-8 text strings interpreted as Unicode scalar sequences. This means this specification defines two control operators: .abnfb for ABNF denoting byte sequences and .abnf for denoting sequences of Unicode scalar values (codepoint) represented as UTF-8 text strings. Both control operators can be applied to targets of either string type; the ABNF is applied to sequence of bytes in the string interpreting that as a sequence of bytes (.abnfb) or as a sequence of code points represented as an UTF-8 text string (.abnf). The controller string MUST be a text string.¶
• ABNF defines a list of rules, not a single expression (called "elements" in [RFC5234]). This is resolved by requiring the controller string to be one valid "element", followed by zero or more valid "rule" separated from the element by a newline; so the controller string can be built by preceding a piece of valid ABNF by an "element" that selects from that ABNF and a newline.¶
• For the same reason, ABNF requires newlines; specifying newlines in CDDL text strings is tedious (and leads to essentially unreadable ABNF). The workaround employs the .cat operator introduced in Section 2.2 and the syntax for text in byte strings. As is customary for ABNF, the syntax of ABNF itself (NOT the syntax expressed in ABNF!) is relaxed to allow a single linefeed as a newline:¶

   CRLF = %x0A / %x0D.0A

• One set of rules provided in an ABNF specification is often used in multiple positions, in particular staples such as DIGIT and ALPHA. (Note that all rules referenced need to be defined in each ABNF operator controller string -- there is no implicit import of [RFC5234] Core ABNF or other rules.) The composition this calls for can be provided by the .cat operator, and/or by .det if there is indentation to be disposed of.¶

These points are combined into an example in Figure 5, which uses ABNF from [RFC3339] to specify one each of the CBOR tags defined in [RFC8943] and [RFC8949].¶

; for RFC 8943
Tag1004 = #6.1004(text .abnf full-date)
; for RFC 8949
Tag0 = #6.0(text .abnf date-time)

full-date = "full-date" .cat rfc3339
date-time = "date-time" .cat rfc3339

; Note the trick of idiomatically starting with a newline, separating
;   off the element in the concatenations above from the rule-list
rfc3339 = '
   date-fullyear   = 4DIGIT
   date-month      = 2DIGIT  ; 01-12
   date-mday       = 2DIGIT  ; 01-28, 01-29, 01-30, 01-31 based on
                             ; month/year
   time-hour       = 2DIGIT  ; 00-23
   time-minute     = 2DIGIT  ; 00-59
   time-second     = 2DIGIT  ; 00-58, 00-59, 00-60 based on leap sec
                             ; rules
   time-secfrac    = "." 1*DIGIT
   time-numoffset  = ("+" / "-") time-hour ":" time-minute
   time-offset     = "Z" / time-numoffset

   partial-time    = time-hour ":" time-minute ":" time-second
                     [time-secfrac]
   full-date       = date-fullyear "-" date-month "-" date-mday
   full-time       = partial-time time-offset

   date-time       = full-date "T" full-time
' .det rfc5234-core

rfc5234-core = '
   DIGIT          =  %x30-39 ; 0-9
   ; abbreviated here
'

4. Features

Commonly, the kind of validation enabled by languages such as CDDL provides a Boolean result: valid, or invalid.¶

In rapidly evolving environments, this is too simplistic. The data models described by a CDDL specification may continually be enhanced by additional features, and it would be useful even for a specification that does not yet describe a specific future feature to identify the extension point the feature can use, accepting such extensions while marking them as such.¶

The .feature control annotates the target as making use of the feature named by the controller. The latter will usually be a string. A tool that validates an instance against that specification may mark the instance as using a feature that is annotated by the specification.¶

More specifically, the tool's diagnostic output might contain the controller (right-hand side) as a feature name, and the target (left-hand side) as a feature detail. However, in some cases, the target has too much detail, and the specification might want to hint the tool that more limited detail is appropriate. In this case, the controller should be an array, with the first element being the feature name (that would otherwise be the entire controller), and the second element being the detail (usually another string), as illustrated in Figure 6.¶

foo = {
  kind: bar / baz .feature (["foo-extensions", "bazify"])
}
bar = ...
baz = ... ; complex stuff that doesn't all need to be in the detail

Figure 7 shows what could be the definition of a person, with potential extensions beyond name and organization being marked further-person-extension. Extensions that are known at the time this definition is written can be collected into $$person-extensions. However, future extensions would be deemed invalid unless the wildcard at the end of the map is added. These extensions could then be specifically examined by a user or a tool that makes use of the validation result; the label (map key) actually used makes a fine feature detail for the tool's diagnostic output.¶

Leaving out the entire extension point would mean that instances that make use of an extension would be marked as whole-sale invalid, making the entire validation approach much less useful. Leaving the extension point in, but not marking its use as special, would render mistakes such as using the label "organisation" instead of "organization" invisible.¶

person = {
  ? name: text
  ? organization: text
  $$person-extensions
  * (text .feature "further-person-extension") => any
}

$$person-extensions //= (? bloodgroup: text)

Figure 8 shows another example where .feature provides for type extensibility.¶

allowed-types = number / text / bool / null
              / [* number] / [* text] / [* bool]
              / (any .feature "allowed-type-extension")

A CDDL tool may simply report the set of features being used; the control then only provides information to the process requesting the validation. One could also imagine a tool that takes arguments allowing the tool to accept certain features and reject others (enable/disable). The latter approach could for instance be used for a JSON/CBOR switch, as illustrated in Figure 9, using SenML [RFC8428] as the example data model used with both JSON and CBOR.¶

SenML-Record = {
; ...
  ? v => number
; ...
}
v = JC<"v", 2>
JC<J,C> = J .feature "json" / C .feature "cbor"

It remains to be seen if the enable/disable approach can lead to new idioms of using CDDL. The language currently has no way to enforce mutually exclusive use of features, as would be needed in this example.¶

5. IANA Considerations

This document requests IANA to register the contents of Table 2 into the registry "CDDL Control Operators" of [IANA.cddl]:¶

6. Implementation Status

This section is to be removed before publishing as an RFC.¶

An early implementation of the control operator .feature has been available in the CDDL tool described in Appendix F of [RFC8610] since version 0.8.11. The validator warns about each feature being used and provides the set of target values used with the feature. The other control operators defined in this specification are also implemented as of version 0.8.21 and 0.8.26 (double-handed .det).¶

Andrew Weiss' [CDDL-RS] has an ongoing implementation of this draft which is feature-complete except for the ABNF and dedenting support (https://github.com/anweiss/cddl/pull/79).¶

7. Security considerations

The security considerations of [RFC8610] apply.¶

While both [RFC5234] and [RFC7405] state that security is truly believed to be irrelevant to the respective document, the use of formal description techniques cannot only simplify, but sometimes also complicate a specification. This can lead to security problems in implementations and in the specification itself. As with CDDL itself, ABNF should be judiciously applied, and overly complex (or "cute") constructions should be avoided.¶

Acknowledgements

Jim Schaad suggested several improvements. The .feature feature was developed out of a discussion with Henk Birkholz. Paul Kyzivat helped isolate the need for .det.¶

.det is an abbreviation for "dedenting cat", but Det is also the name of a German TV Cartoon character created in the 1960s.¶

Author's Address