So what's the proposal?

Make strconv.unquote and strconv.appendQuotedWith public, of course. Possibly with slightly different names.

CC @dsnet

the appendQuotedWith family of functions being exposed would have saved me a few workdays in 2020. IIRC we had to fork and re-export the relevant parts of strconv.

We had this discussion on #45033 and decided not to export unquote. I don't see any new evidence here.

"You don't always want to parse it, but you always need to identify a boundary."

Really?
Not a single time in my not particularly short career I wanted to find out the boundary without also unescaping the string into a usable form (if only for debugging print out). :-)

For whatever reason, various data mining setups still rely very strongly on CSV style files. Huge CSV style files, where there are many strings, escapes abundant and fields are separated by a single ascii codepoint.

By providing a crippled API in the runtime, you're basically saying: don't use it, just go for third-party/self implemented solution. Yep, that's what we do all the time.

But then, why pollute the runtime with useless things?

Based on the discussion above, this proposal seems like a likely decline.
— rsc for the proposal review group

No change in consensus, so declined.
— rsc for the proposal review group