#calibration as child of #progenitor

[Markus Demleitner]

Dear DAL and Semantics,

On Thu, Jun 11, 2020 at 06:09:15PM -0700, Patrick Dowler wrote:
> I always thought of #calibration as meaning "is a #calibration file for
> #this". For example, if #this is raw data, the #calibration might be the
> flat field that one could use to calibrate #this.

Oh wow.  Goes to show how important good and precise definitions are.
I'll keep a reference to this thread for when people in VEPs protest
I shouldn't be so picky with the definition.

Anyway, right now, #calibration's definition says:

  resource used to calibrate the primary data

Note the past tense, which would make, I think, people not lean
towards Pat's interpretation as things are.

> If that raw #this also had a #derivation link to the calibrated product,
> that thing would list both the raw data and the calibration as #progenitor,
> and it seems like trying to say something like "link to the calibration
> progenitor" with a single semantics term is distinguishing what I think are
> roles in Prov DM.

I'd like this outcome just as much as pulling #calibration under
#progenitor.

> In general, I think the DataLink style terms must be thought of with "is"
> and not "was" in front. "was a calibration" is equivalent to putting
> calibration under progenitor... I'm feeling really pedantic here, but it

I'm not really sure tense in the definitions helps a lot; "is a
#progenitor" and "was a #progenitor" in the end doesn't make a lot of
difference semantically, I guess.

Be that as it may, we need to get the interpretation straight, and
while I'm fine with both ways forward, I'd like to pick one.  So, the
VEP could say one of

(1) #calibration becomes a child of #progenitor; if we want terms for
calibration files going with raw data, we'd need new ones (my
original proposal).

(2) We'd fix the descriptions of #calibration, #bias, #dark, and
#flat along the lines of

  #calibration: Data that can be used in the reduction of #this.
    To document calibration data when #this is already reduced,
    use #progenitor (or a child of it).
  #flat: A flat field or similar data on the raw response of the
    instrument used to obtain #this.  Only use this if the item has
    not already been applied to #this.

This means that there is no way to specifically say "this is the
flatfield we've used to reduce #this"; does anyone want this at this
point?


Since Pat prefers (2) and I suspect CADC are the main users of the
#calibration children, I'd go for (2) unless people protest.

So... Do you use any of #bias, #dark, #flat, or #calibration in your
datalink services or know of any such use?  If so, is that compatible
with (2)?

             -- Markus