The CSV spec (RFC 4180) identifies header and charset parameters, but it strangely doesn't specify where they go, or how they're encoded. For header it says:
The "header" parameter indicates the presence or absence of the header line. Valid values are "present" or "absent".
But where does this go in a CSV file or stream? And where does charset go?
See section 3 of the spec for the parameters bit. The whole doc is only a couple of pages of actual content, as we might expect for a CSV spec, but it leaves out critical details. Note that I'm not asking about a web server-to-browser scenario where there'd be HTTP header fields for MIME type or charset, but rather the general case of CSV files and streams.
Confirming that there really is nowhere in a CSV data file to put those metadata fields, either as a file or as a stream.
Those two fields are - as you mentioned - only for cases where a CSV file is being exchanged via a MIME aware protocol, such as HTTP.
Other formats have found different ways around the problem. E.g. JSON must be encoded using UTF-8. XML has an encoding in the prolog.