The spec should say something about character encoding. The default encoding in FIX tag=value is US-ASCII, so converting it to JSON will implicitly be in that encoding. However, for web APIs, the expectation would be to use UTF-8 by default.

The spec mentions the limitations of the JavaScript Date object, but millisecond precision will be insufficient for some users. MiFID II requires microsecond or nanosecond precision, so the GTC has developed Time Precision Enhancement Technical Addendum that covers this topic.

• Convert document to markdown
• Separate into multiple files
• Document command to compile separate files into one document

The spec currently says

The name of a repeating group is the name of the associated NumInGroup field name as it appears in the FIX repository.

Why not use the name of the repeating group from the repository rather than its associated NumInGrp counter? Especially since JSON encoding doesn't need the array counter.

The example gives NoMDEntries as the repeating group object name. The repository gives two repeating group components that use that same counter, MDFullGrp and MDIncGrp. It should be one of those names to distinguish which group.

The first release candidate of FIX JSON primarily considered the use case of translating tag=value encoding to JSON for web presentation of FIX messages. Such translation from existing messages does not require a formalized schema.

However, JSON has been popularized for web APIs as, largely as a substitute for complicated XML APIs. For a potential RC2, this use case should be considered. For example, a REST API could be provided that filters data by FIX fields encoded in JSON, while the result returns JSON-encoded FIX messages.