I hope to find time to prototype it in https://www.oilshell.org/ first. Although it's very simple -- about as simple as JSON -- I think it needs some real world use first to prove it.
I think adding metadata is orthogonal to cleaning up the problems with TSV, e.g. that you can't represent strings with tabs (or newlines) in fields. Also I wouldn't want a TSV enhancement to depend on YAML, as it's quite a big and confusing (e.g. https://news.ycombinator.com/item?id=17359376)
I think it depends on how YAML is used. Requiring full YAML on an input is cumbersome on users (e.g. if any downstream user MUST parse ANY YAML). On the other hand, a simple format which happens to be YAML-compatible means anyone with a YAML parser can parse it without reinventing wheels. That seems a good thing.
An example of that is JSON. Any YAML parser will parse JSON.
TSVX merely seems to guarantee that the output is a simplified (but compatible) form of YAML. All this means is that any YAML parser will read it, and you don't need to write your own. But it's just a colon-delimited dictionary with a few constraints.
FWIW I have my own TSV enhancement that is more modest:
https://github.com/oilshell/oil/wiki/TSV2-Proposal
I hope to find time to prototype it in https://www.oilshell.org/ first. Although it's very simple -- about as simple as JSON -- I think it needs some real world use first to prove it.
I think adding metadata is orthogonal to cleaning up the problems with TSV, e.g. that you can't represent strings with tabs (or newlines) in fields. Also I wouldn't want a TSV enhancement to depend on YAML, as it's quite a big and confusing (e.g. https://news.ycombinator.com/item?id=17359376)