My understanding is that John Parejko's work will most likely make this issue obsolete, but that we don't yet have an alternative system. So I'm not sure we need to nail down more specific requirements. That said...
Is this ticket trying to ingest calibrations that we have created ourselves (e.g. in pipe_drivers or cp_pipe) or are these external calibrations (like DECam community pipeline)?
The original intent was "both".
Do we expect sufficient metadata to be present in the files or are we providing extra metadata on the commandline?
My meaning with "without providing any information... that was automatically inferred in Gen 2" was that we should not narrow the definition of "sufficient metadata" in going from Gen 2 to Gen 3. I would of course prefer that everything be automated, but I take it that there is Gen 3 metadata that doesn't have a Gen 2 equivalent?
Do we know which instrument from the files or do we specify it separately? How do we specify it? Which filter?
No requirement (except that AFAIK filter is covered by the previous point, in that it was automatically inferred in Gen 2).
Are we doing metadata extraction in a generic way as for astro_metadata_translator? Which exposures went into the calibration?
I'm not sure I know enough about astro_metadata_translator to answer either of these questions.