3. Creating CBOR-Based Protocols

Data formats such as CBOR are often used in environments where there is no format negotiation. A specific design goal of CBOR is to not need any included or assumed schema: a decoder can take a CBOR item and decode it with no other knowledge.

Of course, in real-world implementations, the encoder and the decoder will have a shared view of what should be in a CBOR data item. For example, an agreed-to format might be "the item is an array whose first value is a UTF-8 string, second value is an integer, and subsequent values are zero or more floating-point numbers" or "the item is a map that has byte strings for keys and contains at least one pair whose key is 0xab01".

This specification puts no restrictions on CBOR-based protocols. An encoder can be capable of encoding as many or as few types of values as is required by the protocol in which it is used; a decoder can be capable of understanding as many or as few types of values as is required by the protocols in which it is used. This lack of restrictions allows CBOR to be used in extremely constrained environments.

This section discusses some considerations in creating CBOR-based protocols. It is advisory only and explicitly excludes any language from RFC 2119 other than words that could be interpreted as "MAY" in the sense of RFC 2119.

3.1. CBOR in Streaming Applications

In a streaming application, a data stream may be composed of a sequence of CBOR data items concatenated back-to-back. In such an environment, the decoder immediately begins decoding a new data item if data is found after the end of a previous data item.

Not all of the bytes making up a data item may be immediately available to the decoder; some decoders will buffer additional data until a complete data item can be presented to the application. Other decoders can present partial information about a top-level data item to an application, such as the nested data items that could already be decoded, or even parts of a byte string that hasn't completely arrived yet.

Note that some applications and protocols will not want to use indefinite-length encoding. Using indefinite-length encoding allows an encoder to not need to marshal all the data for counting, but it requires a decoder to allocate increasing amounts of memory while waiting for the end of the item. This might be fine for some applications but not others.

3.2. Generic Encoders and Decoders

A generic CBOR decoder can decode all well-formed CBOR data and present them to an application. CBOR data is well-formed if it uses the initial bytes, as well as the byte strings and/or data items that are implied by their values, in the manner defined by CBOR, and no extraneous data follows (Appendix C).

Even though CBOR attempts to minimize these cases, not all well-formed CBOR data is valid: for example, the format excludes simple values below 32 that are encoded with an extension byte. Also, specific tags may make semantic constraints that may be violated, such as by including a tag in a bignum tag or by following a byte string within a date tag. Finally, the data may be invalid, such as invalid UTF-8 strings or date strings that do not conform to [RFC3339]. There is no requirement that generic encoders and decoders make unnatural choices for their application interface to enable the processing of invalid data. Generic encoders and decoders are expected to forward simple values and tags even if their specific codepoints are not registered at the time the encoder/decoder is written (Section 3.5).

Generic decoders provide ways to present well-formed CBOR values, both valid and invalid, to an application. The diagnostic notation (Section 6) may be used to present well-formed CBOR values to humans.

Generic encoders provide an application interface that allows the application to specify any well-formed value, including simple values and tags unknown to the encoder.

3.3. Syntax Errors

A decoder encountering a CBOR data item that is not well-formed generally can choose to completely fail the decoding (issue an error and/or stop processing altogether), substitute the problematic data and data items using a decoder-specific convention that clearly indicates there has been a problem, or take some other action.

3.3.1. Incomplete CBOR Data Items

The representation of a CBOR data item has a specific length, determined by its initial bytes and by the structure of any data items enclosed in the data items. If less data is available, this can be treated as a syntax error. A decoder may also implement incremental parsing, that is, decode the data item as far as it is available and present the data found so far (such as in an event-based interface), with the option of continuing the decoding once further data is available.

Examples of incomplete data items include:

• A decoder expects a certain number of array or map entries but instead encounters the end of the data.

• A decoder processes what it expects to be the last pair in a map and comes to the end of the data.

• A decoder has just seen a tag and then encounters the end of the data.

• A decoder has seen the beginning of an indefinite-length item but encounters the end of the data before it sees the "break" stop code.

3.3.2. Malformed Indefinite-Length Items

Examples of malformed indefinite-length data items include:

• Within an indefinite-length byte string or text, a decoder finds an item that is not of the appropriate major type before it finds the "break" stop code.

• Within an indefinite-length map, a decoder encounters the "break" stop code immediately after reading a key (the value is missing).

Another error is finding a "break" stop code at a point in the data where there is no immediately enclosing (unclosed) indefinite-length item.

3.3.3. Unknown Additional Information Values

At the time of writing, some additional information values are unassigned and reserved for future versions of this document (see Section 5.2). Since the overall syntax for these additional information values is not yet defined, a decoder that sees an additional information value that it does not understand cannot continue parsing.

3.4. Other Decoding Errors

A CBOR data item may be syntactically well-formed but present a problem with interpreting the data encoded in it in the CBOR data model. Generally speaking, a decoder that finds a data item with such a problem might issue a warning, might stop processing altogether, might handle the error and make the problematic value available to the application as such, or take some other type of action.

Such problems might include:

Duplicate keys in a map: Generic decoders (Section 3.2) make data available to applications using the native CBOR data model. That data model includes maps (key-value mappings with unique keys), not multimaps (key-value mappings where multiple entries can have the same key). Thus, a generic decoder that gets a CBOR map item that has duplicate keys will decode to a map with only one instance of that key, or it might stop processing altogether. On the other hand, a "streaming decoder" may not even be able to notice (Section 3.7).

Inadmissible type on the value following a tag: Tags (Section 2.4) specify what type of data item is supposed to follow the tag; for example, the tags for positive or negative bignums are supposed to be put on byte strings. A decoder that decodes the tagged data item into a native representation (a native big integer in this example) is expected to check the type of the data item being tagged. Even decoders that don't have such native representations available in their environment may perform the check on those tags known to them and react appropriately.

Invalid UTF-8 string: A decoder might or might not want to verify that the sequence of bytes in a UTF-8 string (major type 3) is actually valid UTF-8 and react appropriately.

3.5. Handling Unknown Simple Values and Tags

A decoder that comes across a simple value (Section 2.3) that it does not recognize, such as a value that was added to the IANA registry after the decoder was deployed or a value that the decoder chose not to implement, might issue a warning, might stop processing altogether, might handle the error by making the unknown value available to the application as such (as is expected of generic decoders), or take some other type of action.

A decoder that comes across a tag (Section 2.4) that it does not recognize, such as a tag that was added to the IANA registry after the decoder was deployed or a tag that the decoder chose not to implement, might issue a warning, might stop processing altogether, might handle the error and present the unknown tag value together with the contained data item to the application (as is expected of generic decoders), might ignore the tag and simply present the contained data item only to the application, or take some other type of action.

3.6. Numbers

For the purposes of this specification, all number representations for the same numeric value are equivalent. This means that an encoder can encode a floating-point value of 0.0 as the integer 0. It, however, also means that an application that expects to find integer values only might find floating-point values if the encoder decides these are desirable, such as when the floating-point value is more compact than a 64-bit integer.

An application or protocol that uses CBOR might restrict the representations of numbers. For instance, a protocol that only deals with integers might say that floating-point numbers may not be used and that decoders of that protocol do not need to be able to handle floating-point numbers. Similarly, a protocol or application that uses CBOR might say that decoders need to be able to handle either type of number.

CBOR-based protocols should take into account that different language environments pose different restrictions on the range and precision of numbers that are representable. For example, the JavaScript number system treats all numbers as floating point, which may result in silent loss of precision in decoding integers with more than 53 significant bits. A protocol that uses numbers should define its expectations on the handling of non-trivial numbers in decoders and receiving applications.

A CBOR-based protocol that includes floating-point numbers can restrict which of the three formats (half-precision, single-precision, and double-precision) are to be supported. For an integer-only application, a protocol may want to completely exclude the use of floating-point values.

A CBOR-based protocol designed for compactness may want to exclude specific integer encodings that are longer than necessary for the application, such as to save the need to implement 64-bit integers. There is an expectation that encoders will use the most compact integer representation that can represent a given value. However, a compact application should accept values that use a longer-than-needed encoding (such as encoding "0" as 0b000_11101 followed by two bytes of 0x00) as long as the application can decode an integer of the given size.

3.7. Specifying Keys for Maps

The encoding and decoding applications need to agree on what types of keys are going to be used in maps. In applications that need to interwork with JSON-based applications, keys probably should be limited to UTF-8 strings only; otherwise, there has to be a specified mapping from the other CBOR types to Unicode characters, and this often leads to implementation errors. In applications where keys are numeric in nature and numeric ordering of keys is important to the application, directly using the numbers for the keys is useful.

If multiple types of keys are to be used, consideration should be given to how these types would be represented in the specific programming environments that are to be used. For example, in JavaScript objects, a key of integer 1 cannot be distinguished from a key of string "1". This means that, if integer keys are used, the simultaneous use of string keys that look like numbers needs to be avoided. Again, this leads to the conclusion that keys should be of a single CBOR type.

Decoders that deliver data items nested within a CBOR data item immediately on decoding them ("streaming decoders") often do not keep the state that is necessary to ascertain uniqueness of a key in a map. Similarly, an encoder that can start encoding data items before the enclosing data item is completely available ("streaming encoder") may want to reduce its overhead significantly by relying on its data source to maintain uniqueness.

A CBOR-based protocol should make an intentional decision about what to do when a receiving application does see multiple identical keys in a map. The resulting rule in the protocol should respect the CBOR data model: it cannot prescribe a specific handling of the entries with the identical keys, except that it might have a rule that having identical keys in a map indicates a malformed map and that the decoder has to stop with an error. Duplicate keys are also prohibited by CBOR decoders that are using strict mode (Section 3.10).

The CBOR data model for maps does not allow ascribing semantics to the order of the key/value pairs in the map representation. Thus, it would be a very bad practice to define a CBOR-based protocol in such a way that changing the key/value pair order in a map would change the semantics, apart from trivial aspects (cache usage, etc.). (A CBOR-based protocol can prescribe a specific order of serialization, such as for canonicalization.)

Applications for constrained devices that have maps with 24 or fewer frequently used keys should consider using small integers (and those with up to 48 frequently used keys should consider also using small negative integers) because the keys can then be encoded in a single byte.

3.8. Undefined Values

In some CBOR-based protocols, the simple value (Section 2.3) of Undefined might be used by an encoder as a substitute for a data item with an encoding problem, in order to allow the rest of the enclosing data items to be encoded without harm.

3.9. Canonical CBOR

Some protocols may want encoders to only emit CBOR in a particular canonical format; those protocols might also have the decoders check that their input is canonical. Those protocols are free to define what they mean by a canonical format and what encoders and decoders are expected to do. This section lists some suggestions for such protocols.

If a protocol considers "canonical" to mean that two encoder implementations starting with the same input data will produce the same CBOR output, the following four rules would suffice:

• Integers must be as small as possible.

  • 0 to 23 and -1 to -24 must be expressed in the same byte as the major type;

  • 24 to 255 and -25 to -256 must be expressed only with an additional uint8_t;

  • 256 to 65535 and -257 to -65536 must be expressed only with an additional uint16_t;

  • 65536 to 4294967295 and -65537 to -4294967296 must be expressed only with an additional uint32_t.

• The expression of lengths in major types 2 through 5 must be as short as possible. The rules for these lengths follow the above rule for integers.

• The keys in every map must be sorted lowest value to highest. Sorting is performed on the bytes of the representation of the key data items without paying attention to the 3/5 bit splitting for major types. (Note that this rule allows maps that have keys of different types, even though that is probably a bad practice that could lead to errors in some canonicalization implementations.) The sorting rules are:

  • If two keys have different lengths, the shorter one sorts earlier;

  • If two keys have the same length, the one with the lower value in (byte-wise) lexical order sorts earlier.

• Indefinite-length items must be made into definite-length items.

If a protocol allows for IEEE floats, then additional canonicalization rules might need to be added. One example rule might be to have all floats start as a 64-bit float, then do a test conversion to a 32-bit float; if the result is the same numeric value, use the shorter value and repeat the process with a test conversion to a 16-bit float. (This rule selects 16-bit float for positive and negative Infinity as well.) Also, there are many representations for NaN. If NaN is an allowed value, it must always be represented as 0xf97e00.

CBOR tags present additional considerations for canonicalization. The absence or presence of tags in a canonical format is determined by the optionality of the tags in the protocol. In a CBOR-based protocol that allows optional tagging anywhere, the canonical format must not allow them. In a protocol that requires tags in certain places, the tag needs to appear in the canonical format. A CBOR-based protocol that uses canonicalization might instead say that all tags that appear in a message must be retained regardless of whether they are optional.

3.10. Strict Mode

Some areas of application of CBOR do not require canonicalization (Section 3.9) but may require that different decoders reach the same (semantically equivalent) results, even in the presence of potentially malicious data. This can be required if one application (such as a firewall or other protecting entity) makes a decision based on the data that another application, which independently decodes the data, relies on.

Normally, it is the responsibility of the sender to avoid ambiguously decodable data. However, the sender might be an attacker specially making up CBOR data such that it will be interpreted differently by different decoders in an attempt to exploit that as a vulnerability. Generic decoders used in applications where this might be a problem need to support a strict mode in which it is also the responsibility of the receiver to reject ambiguously decodable data. It is expected that firewalls and other security systems that decode CBOR will only decode in strict mode.

A decoder in strict mode will reliably reject any data that could be interpreted by other decoders in different ways. It will reliably reject data items with syntax errors (Section 3.3). It will also expend the effort to reliably detect other decoding errors (Section 3.4). In particular, a strict decoder needs to have an API that reports an error (and does not return data) for a CBOR data item that contains any of the following:

• a map (major type 5) that has more than one entry with the same key

• a tag that is used on a data item of the incorrect type

• a data item that is incorrectly formatted for the type given to it, such as invalid UTF-8 or data that cannot be interpreted with the specific tag that it has been tagged with

A decoder in strict mode can do one of two things when it encounters a tag or simple value that it does not recognize:

• It can report an error (and not return data).

• It can emit the unknown item (type, value, and, for tags, the decoded tagged data item) to the application calling the decoder with an indication that the decoder did not recognize that tag or simple value.

The latter approach, which is also appropriate for non-strict decoders, supports forward compatibility with newly registered tags and simple values without the requirement to update the encoder at the same time as the calling application. (For this, the API for the decoder needs to have a way to mark unknown items so that the calling application can handle them in a manner appropriate for the program.)

Since some of this processing may have an appreciable cost (in particular with duplicate detection for maps), support of strict mode is not a requirement placed on all CBOR decoders.

Some encoders will rely on their applications to provide input data in such a way that unambiguously decodable CBOR results. A generic encoder also may want to provide a strict mode where it reliably limits its output to unambiguously decodable CBOR, independent of whether or not its application is providing API-conformant data.