| Internet-Draft | Packed CBOR: Table set up by reference | March 2024 |
| Amsüss | Expires 5 September 2024 | [Page] |
Packed CBOR is a compression mechanism for Concise Binary Object Representation (CBOR) that can be used without a decompression step. This document introduces a means for setting up its tables by means of dereferencable identifiers, and introduces a pattern of using it without sending long identifiers.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://chrysn.codeberg.page/packed-by-reference/draft-amsuess-cbor-packed-by-reference.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-amsuess-cbor-packed-by-reference/.¶
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://codeberg.org/chrysn/packed-by-reference.¶
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 5 September 2024.¶
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.¶
[ See abstract. ]¶
CBOR tag TBD114 is defined with semantics similar to tags TBD113 and TBD1113 from [I-D.ietf-cbor-packed] in that it sets up tables around a rump.¶
Packed-By-Reference = #6.<tbd114>([count, source, rump]) rump = any source = CRI / ~uri count = (count-shared-and-argument // ; similar to tag 113 count-shared, count-argument ) ; similar to tag 1113 count-shared-and-argument = uint count-shared = uint count-argument = uint tbd114 = 114 ; preliminary value, see IANA considerations¶
The items inserted by the tables are not given explicitly,
but picked out of tables known by their identifier given as source.
Such a source needs to represent two lists of CBOR items,
one for each kind of tables (one for shared item, one for arguments).
The tag prepends some number of items out of those source lists
to the tables that are used to decompress the rump.¶
The identifier is given as a URI string (as defined in [RFC3986]
or equivalently as a CRI (as defined in [I-D.ietf-core-href].
Later iterations of this document may introduce additional options.If the stand-in concept of [I-D.bormann-cbor-yang-standin] is generalized,
the source item may become the raw list of tables,
possibly disallowing the CRI and URI variants.
Given that tags 113 and 1113 are capable of expressing cases where the source tables are present,
tag TBD114 should then be used by using a dereferencing stand-in in the source position.
When the source identifier is dereferencable,
all considerations from [I-D.bormann-t2trg-deref-id] apply.
(Simplifying: No dereferencing at runtime -- the recipient either knows it already or treats it as unknown).¶
If the same number of items is prepended to both tables, their count is given as a single number; otherwise, the numbers are given separately.¶
Encoders SHOULD use the most compact form of count, and SHOULD pick the lowest count(s) sufficient to encode the items contained in the rump. When those conflict, they may priorize either. If the source supports evolution of sources (see Section 2.1.2), disregarding that recommendation may pose an interoperability hazard.¶
The count encoded for the number of table entries given in a document will often mismatch with the number of entries the receiver of a document knows to be present in the given source.¶
If the encoded count is less than the number of known entries, this merely expresses that the originator of the document did not use the higher numbers. When a document's tables are populated from multiple sources, encoding the smallest possible count is useful because the table indices used throughout the document stay small and can thus be encoded concisely.¶
If the encoded count is larger than the number of known entries, this indicates that the document may contain references that the receiver does not know. This can happen when a source has been evolved compatibly to contain more entries, compared to when the receiver learned of the source definition. Source entries beyond the receiver's knowledge stay unpopulated in the receiver's tables, but still shift existing entries to higher indices.¶
Some CBOR protocols with elements that support isolation of processing errors. For example, a CRI that uses unknown extensions is regarded as "unprocessable" (Section 5.2.1 of [I-D.ietf-core-href]). It cannot be resolved, is unequal to any other CRI (unless they are identical), but does not inhibit the processing of its surrounding document.¶
In such protocols, references to unpopulated table entries can be tolerated as described in Section 2.1 of [I-D.ietf-cbor-packed]. Care has to be taken around processing tag TBD1112: If that tag is produced in the course of unpacking, comparisons for identity are not reliable. Similarly, if the unpacking mechanism provides access to the serialized form of the unprocessable entity, identity comparisons are only reliable if the items being compared have the same table setup applied.¶
Protocols may also pre-populated entries with values that are reserved in the protocol and specified to be ignored at reception. Later, when the entries are specified, concrete values take their places. This has roughly the same effect, but is harder to describe. (This paragraph may be removed later unless it is found to be particularly useful).¶
If a protocol has isolated error processing only around some elements, encoders need to take care to not use entries in unisolated positions that may be unpopulated at the decoder. The protocol and source authors need to provide appropriate guidance.¶
Protocols that do not support error isolation need a way to negotiate the understood set of sources and table entries.¶
If an implementation does not support any elements with isolated error processing at all, a receiver of a document may already stop processing the document when encountering a setup by reference that includes undefined elements.¶
The content of tables may be altered in more ways than just adding entries that were previously unpopulated. Such changes are NOT RECOMMENDED, because while they can be done in a compatible way, providing criteria for it are out of scope of this document.¶
If a later version of this document uses stand-in values more actively, this section will need to be revisited: In that case, the tables may be part of the outer source, and then those would grow internally.¶
If a large number of items at the beginning of the source tables would not be used,
there is an additional four-argument form of count
that defines a number of items in the source tables that are skipped
before selecting items into the table.
This allows keeping the indices low and therefore compact.¶
count //= (
skip-shared, count-shared, skip-argument, count-argument
)
skip-shared = uint
skip-argument = uint
¶
Source tables should be designed in such a way that commonly used items are at the start to avoid the necessity of the four-argument form.¶
Suppose the URI "tag:example.com,2023:byref" defines the items
["price", "category", "author", "title", "fiction", 8.95, "isbn"] in both tables.
Then the example in figure 3 of [I-D.ietf-cbor-packed] can be written as:¶
114([7, "tag:example.com,2023:books"
[{"store": {
"book": [
{simple(1): "reference", simple(2): "Nigel Rees",
simple(3): "Sayings of the Century", simple(0): simple(5)},
{simple(1): simple(4), simple(2): "Evelyn Waugh",
simple(3): "Sword of Honour", simple(0): 12.99},
{simple(1): simple(4), simple(2): "Herman Melville",
simple(3): "Moby Dick", simple(6): "0-553-21311-3",
simple(0): simple(5)},
{simple(1): simple(4), simple(2): "J. R. R. Tolkien",
simple(3): "The Lord of the Rings",
simple(6): "0-395-19395-8", simple(0): 22.99}],
"bicycle": {"color": "red", simple(0): 19.95}}}]])
¶
Assuming that the underlying CBOR protocol defines that unknown keys on goods may be ignored, an older receiver that only knows the first 5 entries of the source tables could still process the document, but would be missing all ISBNs and the price of one item.¶
Documents that use tables from multiple sources can easily spend many bytes on listing source identifiers. A pattern that reduces the verbosity while staying unambiguous are nested table setups, where the outer tables are extended to contain additional identifiers.¶
In this pattern, tables are set up in two stages:¶
The outer stage contains the CRIs or URIs that may later be used as source values. (It may also contain other items). The inner stage is set up using tag TBD114, and the source given is a packed reference.¶
All table inputs can be evolved orthogonally as described in Section 2.1.2. If an unspecified entry is used as a source, the whole source content is considered unspecified.¶
In this example, the initial table set up is provided by the media type, and contains these items:¶
114([5, 6(42) / outer item 100 /, 114([2, 6(45) / outer item 101, currently item 105 /, [ simple(11) / outer item 0, currenlty item 11, "This class has students with the following names" /, simple(0) / item 0 of german-names, "Franz" /, simple(2) / item 0 of english-names, currently item 2, "George" /, simple(1) / item 1 of german-names, "Fritz" /, simple(7) / item 5 of english-names, currently item 7, "Jack" / ]])])¶
Note that a constrained implementation of a decoder may not even have the fully expanded form of the URIs or CRIs available; it may only be capable of using these table entries in the source position and then find the shipped source lists.¶
[ TBD ]¶
From -01 to -02:¶
[ TBD ]¶