Internet-Draft | CBOR tag for extended time | October 2023 |
Bormann, et al. | Expires 9 April 2024 | [Page] |
The Concise Binary Object Representation (CBOR, RFC 8949) is a data format whose design goals include the possibility of extremely small code size, fairly small message size, and extensibility without the need for version negotiation.¶
In CBOR, one point of extensibility is the definition of CBOR tags. RFC 8949 defines two tags for time: CBOR tag 0 (RFC3339 time as a string) and tag 1 (Posix time as int or float). Since then, additional requirements have become known. The present document defines a CBOR tag for time that allows a more elaborate representation of time, as well as related CBOR tags for duration and time period. It is intended as the reference document for the IANA registration of the CBOR tags defined.¶
The present version (-10) addresses comments and WG discussion during the AD review.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-ietf-cbor-time-tag/.¶
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/time-tag.¶
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 9 April 2024.¶
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.¶
The Concise Binary Object Representation (CBOR, [RFC8949]) provides for the interchange of structured data without a requirement for a pre-agreed schema. RFC 8949 defines a basic set of data types, as well as a tagging mechanism that enables extending the set of data types supported via an IANA registry.¶
In CBOR, one point of extensibility is the definition of CBOR tags. RFC 8949 defines two tags for time: CBOR tag 0 (RFC3339 time as a string) and tag 1 (Posix time as int or float). Since then, additional requirements have become known. The present document defines a CBOR tag for time that allows a more elaborate representation of time, as well as related CBOR tags for duration and time period. It is intended as the reference document for the IANA registration of the CBOR tags defined.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
The term "byte" is used in its now customary sense as a synonym for "octet". Where bit arithmetic is explained, this document uses the notation familiar from the programming language C (including C++14's 0bnnn binary literals), except that the operator "**" stands for exponentiation.¶
CBOR diagnostic notation is defined in Section 8 of [RFC8949] and Appendix G of [RFC8610]. A machine-processable model of the data structures defined in this specification is provided throughout the text using the Concise Data Definition Language, CDDL [RFC8610]; Appendix A provides the collected model information.¶
For the time tag, the present specification addresses the following objectives that go beyond the original tags 0 and 1:¶
Additional resolution for epoch-based time (as in tag 1). CBOR tag 1 only provides for integer and up to binary64 floating point representation of times, limiting resolution to approximately microseconds at the time of writing (and progressively becoming worse over time).¶
Indication of timescale. Tags 0 and 1 are for UTC; however, some interchanges are better performed on TAI. Other timescales may be registered once they become relevant (e.g., one of the proposed successors to UTC that might no longer use leap seconds, or a scale based on smeared leap seconds).¶
By incorporating a way to transport [IXDTF] suffix information (Section 3.6, Section 3.7), additional indications can be provided of intents about the interpretation of the time given, in particular for future times. Intents might include information about time zones, daylight savings times, preferred calendar representations, etc.¶
Semantics not covered by this document can be added by registering additional map keys for the map inside the tag, the specification for which is referenced by the registry entry (Section 7.3, Section 3). For example, map keys could be registered for:¶
Direct representation of natural platform time formats. Some platforms use epoch-based time formats that require some computation to convert them into the representations allowed by tag 1; these computations can also lose precision and cause ambiguities. (The present specification does not take a position on whether tag 1 can be "fixed" to include, e.g., Decimal or BigFloat representations. It does define how to use these representations with the extended time format.)¶
Additional tags are defined for durations and periods.¶
An extended time is indicated by CBOR tag 1001, the content of which is a map data item (CBOR major type 5). The map may contain integer (major types 0 and 1) or text string (major type 3) keys, with the value type determined by each specific key. For negative integer keys and text string values of the key, implementations MUST ignore key/value pairs they do not understand. Conversely, for unsigned integer keys, implementations MUST signal as an error key/value pairs they do not understand or implement (these are either "base time" or "critical", see below).¶
The map must contain exactly one unsigned integer key that specifies the "base time", and may also contain one or more negative integer or text-string keys, which may encode supplementary information.¶
Supplementary information may also be provided by additional unsigned integer keys that are explicitly defined to provide supplementary information ("critical"; as these are required to be understood, there can be no confusion with Base Time keys).¶
Negative integer and text string keys always supply supplementary information ("elective", and this will not be explicitly stated below).¶
Supplementary information may include:¶
a higher precision time offset to be added to the Base Time,¶
a reference timescale and epoch different from the default UTC and 1970-01-01¶
information about clock quality parameters, such as source, accuracy, and uncertainty¶
Additional keys can be defined by registering them in the Map Key Registry (Section 7.3). Future keys may add:¶
intent information such as timezone and daylight savings time, and/or possibly positioning coordinates, to express information that would indicate a local time.¶
This document does not define supplementary text keys. A number of both unsigned and negative-integer keys are defined in the following subsections.¶
A formal definition of Tag 1001 in CDDL is:¶
Etime = #6.1001(etime-detailed) etime-framework = { uint => any ; at least one base time * (nint/text) => any ; elective supplementary information * uint => any ; critical supplementary information } etime-detailed = ({ $$ETIME-BASETIME ClockQuality-group * $$ETIME-ELECTIVE * $$ETIME-CRITICAL * ((nint/text) .feature "etime-elective-extension") => any * (uint .feature "etime-critical-extension") => any }) .within etime-framework¶
Key 1 indicates a Base Time value that is exactly like the data item that would be tagged by CBOR tag 1 (Posix time [TIME_T] as int or float). The time value indicated by the value under this key can be further modified by other keys.¶
$$ETIME-BASETIME //= (1: ~time)¶
Keys 4 and 5 indicate a Base Time value and are like key 1, except that the data item is an array as defined for CBOR tag 4 or 5, respectively. This can be used to include a Decimal or Bigfloat epoch-based float [TIME_T] in an extended time.¶
$$ETIME-BASETIME //= (4: ~decfrac) $$ETIME-BASETIME //= (5: ~bigfloat)¶
The keys -3, -6, -9, -12, -15 and -18 indicate additional decimal fractions by giving an unsigned integer (major type 0) and scaling this with the scale factor 1e-3, 1e-6, 1e-9, 1e-12, 1e-15, and 1e-18, respectively (see Table 1). More than one of these keys MUST NOT be present in one extended time data item. These additional fractions are added to a Base Time in seconds [SI-SECOND] indicated by a Key 1, which then MUST also be present and MUST have an integer value.¶
Key | meaning | example usage |
---|---|---|
-3 | milliseconds | Java time |
-6 | microseconds | (old) UNIX time |
-9 | nanoseconds | (new) UNIX time |
-12 | picoseconds | Haskell time |
-15 | femtoseconds | (future) |
-18 | attoseconds | (future) |
$$ETIME-ELECTIVE //= (-3: uint) $$ETIME-ELECTIVE //= (-6: uint) $$ETIME-ELECTIVE //= (-9: uint) $$ETIME-ELECTIVE //= (-12: uint) $$ETIME-ELECTIVE //= (-15: uint) $$ETIME-ELECTIVE //= (-18: uint)¶
Note that these keys have been provided to facilitate representing
pairs of the form second/decimal fraction of a second, as found for
instance in C timespec
(compare Section 7.27.1 of [C]).
When ingesting a timestamp with one of these keys into a type provided
by the target platform, care has to be taken to meet its invariants.
E.g., for C timespec
, the fractional part tv_nsec
needs to be
between 0 inclusive and 109 exclusive, which can be
achieved by also adjusting the base time appropriately.¶
Key -1 is used to indicate a timescale. The value 0 indicates UTC, with the POSIX epoch [TIME_T]; the value 1 indicates TAI, with the PTP (Precision Time Protocol) epoch (1 January 1970 00:00:00 TAI, see [IEEE1588-2019] or [IEEE1588-2008]).¶
$$ETIME-ELECTIVE //= (-1 => $ETIME-TIMESCALE) $ETIME-TIMESCALE /= &(etime-utc: 0) $ETIME-TIMESCALE /= &(etime-tai: 1)¶
If key -1 is not present, timescale value 0 is implied.¶
Additional values can be registered in the Timescale Registry (Section 7.2); values MUST be integers or text strings.¶
(Note that there should be no timescales "GPS" or "NTP" — instead, the time should be converted to TAI or UTC using a single addition or subtraction.)¶
A number of keys are defined to indicate the quality of clock that was used to determine the point in time.¶
The first three are analogous to clock-quality-grouping
in
[RFC8575], which is in turn based on the definitions in
[IEEE1588-2008]; two more are specific to this document.¶
ClockQuality-group = ( ? &(ClockClass: -2) => uint .size 1 ; PTP/RFC8575 ? &(ClockAccuracy: -4) => uint .size 1 ; PTP/RFC8575 ? &(OffsetScaledLogVariance: -5) => uint .size 2 ; PTP/RFC8575 ? &(Uncertainty: -7) => ~time/~duration ? &(Guarantee: -8) => ~time/~duration )¶
Key -2 (ClockClass) can be used to indicate the clock class as per [RFC8575] (which is based on Table 5 in Section 7.6.2.4 of [IEEE1588-2008]; this has updated language as Table 4 in Section 7.6.2.5 of [IEEE1588-2019]). It is defined as a one-byte unsigned integer as that is the range defined there.¶
Key -4 (ClockAccuracy) can be used to indicate the clock accuracy as per [RFC8575] (which is based on Table 6 in Section 7.6.2.5 of [IEEE1588-2008]; additional values have been defined in Table 5 in Section 7.6.2.6 of [IEEE1588-2019]). It is defined as a one-byte unsigned integer as that is the range defined there. The range between 23 and 47 is a slightly distorted logarithmic scale from 1 ps to 1 s in [IEEE1588-2019] (in [IEEE1588-2008] the range was a subset of that, 32 to 47 for 25 ns to 1 s) — see Figure 2; the number 254 is the value to be used if an unknown accuracy needs to be expressed.¶
Key -5 (OffsetScaledLogVariance) can be used to represent the variance exhibited by the clock when it has lost its synchronization with an external reference clock. The details for the computation of this characteristic are defined in Section 7.6.3 of [IEEE1588-2019] and the same section in [IEEE1588-2008].¶
Key -7 (Uncertainty) can be used to represent a known measurement uncertainty for the clock, as a numeric value in seconds or as a duration (Section 4).¶
For this document, uncertainty is defined as in Section 2.2.3 of [GUM]: "parameter, associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand". More specifically, the value for this key represents the extended uncertainty for k = 2, in seconds.¶
Note that the additional information that can be meaningfully provided with the duration that represents an uncertainty is limited, e.g., it is not customary to provide an uncertainty for a duration representing an uncertainty. Implementations are free to reduce an uncertainty (which is already elective) to the information they can process.¶
Key -8 (Guarantee) can be used to represent a stated guarantee for the accuracy of the point in time, as a numeric value in seconds or as a duration (Section 4) representing the maximum allowed deviation from the true value.¶
While such a guarantee is unattainable in theory, existing standards such as [RFC3161] stipulate the representation of such guarantees, and therefore this format provides a way to represent them as well; the time value given is nominally guaranteed to not deviate from the actual time by more than the value of the guarantee, in seconds.¶
Note that the additional information that can be meaningfully provided with the duration that represents a guarantee is limited, e.g., it is not meaningful to provide a guarantee of accuracy for the duration representing a guarantee of accuracy. Implementations are free to reduce a guarantee (which is already elective) to the information they can process.¶
Keys -10 and 10 supply supplementary information, where key 10 is critical.¶
They can be used to provide a hint about the time zone that
would best fit for displaying the time given to humans, using a text
string in the format defined for time-zone-name
or time-numoffset
in [IXDTF].
Key -10 is equivalent to providing this information as an elective
hint, while key 10 provides this information as critical (i.e., it
MUST be used when interpreting the entry with this key).¶
Keys -10 and 10 MUST NOT both be present.¶
$$ETIME-ELECTIVE //= (-10: time-zone-info) $$ETIME-CRITICAL //= (10: time-zone-info) time-zone-info = tstr .abnf ("time-zone-name / time-numoffset" .det IXDTFtz) IXDTFtz = ' time-hour = 2DIGIT ; 00-23 time-minute = 2DIGIT ; 00-59 time-numoffset = ("+" / "-") time-hour ":" time-minute time-zone-initial = ALPHA / "." / "_" time-zone-char = time-zone-initial / DIGIT / "-" / "+" time-zone-part = time-zone-initial *13(time-zone-char) ; but not "." or ".." time-zone-name = time-zone-part *("/" time-zone-part) ALPHA = %x41-5A / %x61-7A ; A-Z / a-z DIGIT = %x30-39 ; 0-9 ' ; extracted from [IXDTF] and [RFC3339]; update as needed¶
Keys -11 and 11 supply supplementary information, where key 11 is critical.¶
Similar to keys -10 and 10, keys -11 (elective) and 11 (critical) can
be used to provide additional information in the style of IXDTF
suffixes, such as the calendar that would best fit for displaying the
time given to humans.
The key's value is a map that has IXDTF suffix-key
names as keys and
corresponding suffix values as values, specifically:¶
$$ETIME-ELECTIVE //= (-11: suffix-info-map) $$ETIME-CRITICAL //= (11: suffix-info-map) suffix-info-map = { * suffix-key => suffix-values } suffix-key = tstr .abnf ("suffix-key" .det IXDTF) suffix-values = one-or-more<suffix-value> one-or-more<T> = T / [ 2* T ] suffix-value = tstr .abnf ("suffix-value" .det IXDTF) IXDTF = ' key-initial = lcalpha / "_" key-char = key-initial / DIGIT / "-" suffix-key = key-initial *key-char suffix-value = 1*alphanum alphanum = ALPHA / DIGIT lcalpha = %x61-7A ALPHA = %x41-5A / %x61-7A ; A-Z / a-z DIGIT = %x30-39 ; 0-9 ' ; extracted from [IXDTF]; update as needed!¶
When keys -11 and 11 both are present, the two maps MUST NOT have entries with the same map keys.¶
Figure 4 of [IXDTF] gives an example for an extended date-time with both time zone and suffix information:¶
1996-12-19T16:39:57-08:00[America/Los_Angeles][u-ca=hebrew]¶
A time tag that is approximating this example, in CBOR diagnostic notation, would be:¶
/ 1996-12-19T16:39:57-08:00[America//Los_Angeles][u-ca=hebrew] / 1001({ 1: 851042397, -10: "America/Los_Angeles", -11: { "u-ca": "hebrew" } })¶
Note that both -10 and -11 are using negative keys and therefore
provide elective information, as in the IXDTF form.
Note also that in this example the time numeric offset (-08:00
) is
lost in translating from the [RFC3339] information in the IXDTF into a
Posix time that can be included under Key 1 in a time tag.¶
A duration is the length of an interval of time. Durations in this format are given in SI seconds, possibly adjusted for conventional corrections of the timescale given (e.g., leap seconds).¶
Except for using Tag 1002 instead of 1001, durations are structurally identical to time values.¶
Duration = #6.1002(etime-detailed)¶
Semantically, they do not measure the time elapsed from a given epoch, but from the start to the end of (an otherwise unspecified) interval of time.¶
In combination with an epoch identified in the context, a duration can also be used to express an absolute time.¶
Without such context, durations are subject to some uncertainties underlying the timescale used. E.g., for durations intended as a determinant of future time periods, there is some uncertainty of what irregularities (such as leap seconds, timescale corrections) will be exhibited by the timescale in that period. For durations as measurements of past periods, abstracting the period to a duration loses some detail about timescale irregularities. For many applications, these uncertainties are acceptable and thus the use of durations is appropriate.¶
A period is a specific interval of time, specified as either two times giving the start and the end of that interval, or as one of these two plus a duration.¶
They are given as an array of unwrapped time and duration elements, tagged with Tag 1003:¶
Period = #6.1003([ start: ~Time / null end: ~Time / null ? duration: ~Duration / null ])¶
If the third array element is not given, the duration element is null. Exactly two out of the three elements must be non-null, this can be somewhat verbosely expressed in CDDL as:¶
clumsy-Period = #6.1003([ (start: ~Time, ((end: ~Time, ? duration: null) // (end: null, duration: ~Duration))) // (start: null, end: ~Time, duration: ~Duration) ])¶
When detailed validation is not needed, the type names defined in Figure 3 are recommended:¶
RFC Editor: please replace RFCthis with the RFC number of this RFC, and remove this note.¶
This specification defines a new registry titled "Timescales" in the "CBOR Tags" registry group [IANA.cbor-tags], with a combination of "Expert Review" and "RFC Required" as the Registration Procedure (Sections 4.5 and 4.7 of [BCP26]).¶
Each entry needs to provide a timescale name (a sequence of uppercase
ASCII characters and digits, where a digit may not occur at the start:
[A-Z][A-Z0-9]*
), a value (CBOR unsigned integer, uint), and brief description
of the semantics, and a specification reference (RFC).
The initial contents are shown in Table 3.¶
Timescale | Value | Semantics | Reference |
---|---|---|---|
UTC | 0 | UTC with POSIX Epoch | [RFCthis] |
TAI | 1 | TAI with PTP Epoch | [RFCthis] |
This specification defines a new registry titled "Time Tag Map Keys" in the "CBOR Tags" registry group [IANA.cbor-tags], with "Specification Required" as the Registration Procedure (Section 4.6 of [BCP26]).¶
The designated expert is requested to assign the key values with the shortest encodings (1+0 and 1+1 encoding) to registrations that are likely to enjoy wide use and can benefit from short encodings.¶
Each entry needs to provide a map key value (CBOR integer, int), a brief description of the semantics, and a specification reference (RFC). The initial contents are shown in Table 4.¶
Value | Semantics | Reference |
---|---|---|
-18 | attoseconds | [RFCthis] |
-15 | femtoseconds | [RFCthis] |
-12 | picoseconds | [RFCthis] |
-11 | IXDTF Suffix Information (elective) | [RFCthis], [IXDTF] |
-10 | IXDTF Time Zone Hint (elective) | [RFCthis], [IXDTF] |
-9 | nanoseconds | [RFCthis] |
-8 | Guarantee | [RFCthis] |
-7 | Uncertainty | [RFCthis] |
-6 | microseconds | [RFCthis] |
-5 | Offset-Scaled Log Variance | [RFCthis] |
-4 | Clock Accuracy | [RFCthis] |
-3 | milliseconds | [RFCthis] |
-2 | Clock Class | [RFCthis] |
1 | Base Time value as in CBOR Tag 1 | [RFC8949] [RFCthis] |
4 | Base Time value as in CBOR Tag 4 | [RFC8949] [RFCthis] |
5 | Base Time value as in CBOR Tag 5 | [RFC8949] [RFCthis] |
10 | IXDTF Time Zone Hint (critical) | [RFCthis], [IXDTF] |
11 | IXDTF Suffix Information (critical) | [RFCthis], [IXDTF] |
The security considerations of RFC 8949 apply; the tags introduced here are not expected to raise security considerations beyond those.¶
Time, of course, has significant security considerations; these include the exploitation of ambiguities where time is security relevant (e.g., for freshness or in a validity span) or the disclosure of characteristics of the emitting system (e.g., time zone, or clock resolution and wall clock offset).¶
This appendix collects the CDDL rules spread over the document into one convenient place.¶
The authors would like to acknowledge the many comments from members of the CBOR WG, as well as Francesca Palombini for her AD review.¶