Valid and recommended format for xml:lang attribute?

[stefanjakobsson]

The schema documentation for the xml:lang attribute states that allowed values are "IETF BCP 47, ISO 639-1 language codes, e.g. en, de, fr". This seems a bit contradictory since BCP 47, as far as I can tell, allows a lot more than just ISO 639-1. Is everything allowed by BCP 47 valid?

Related to the above: is there a recommended format? We are having some problems with https://citation.crosscite.org/ not picking up the english title of our datasets, and I was wondering if this is due to the fact that we are using "en" (ISO 630-1) instead of "en-US" or "en-GB".