Should SDKs support bytes attributes or not?

[adriangb]

Currently the spec reads:

The attribute value is either:

• A primitive type: string, boolean, double precision floating point (IEEE 754-1985) or signed 64 bit integer.
• An array of primitive type values. The array MUST be homogeneous, i.e., it MUST NOT contain values of different types.

Notably it does not include a bytes/binary value.

But later on in the same document:

Byte sequences (e.g. Go's []byte slice or raw byte content of a file) SHOULD be converted to AnyValue's bytes_value field.

Which seems contradictory?

This came from discussion in open-telemetry/opentelemetry-rust#2471 (comment)

[pellared]

The spec says:

The standard attribute definition SHOULD be used to represent attributes in data modeling unless there is a strong justification to diverge. For example, the Log Data Model has an extended attributes definition allowing values of type Any. This reflects that LogRecord attributes are expected to model data produced from external log APIs, which do not necessarily have the same value type restrictions as the standard attribute definition.

Note: Extending the set of standard attribute value types is a breaking change. This was decided after extensive debate, with arguments as follows:

Therefore bytes are only allowed in log attributes.

[adriangb]

Should the section I linked to be amended then?

[pellared]

SGTM. Feel free to sumbit a PR if you have an idea how to do it.