On Thu, Sep 20, 2012 at 10:18 AM, David Berry <[log in to unmask]> wrote:
> On 20 September 2012 17:50, Tim Jenness <[log in to unmask]> wrote:
>> I'm still not convinced that REF should turn up in provenance because
>> no pixel data from REF contributes to the output.
>
> All depends on what you mean by a parent. I suppose several
> definitions are possible. Which is the most useful is not clear to me.
> We could change the defining NDF event from "opened for read" to
> "mapped for read" if necessary.
>
I tend to think that a parent is something that contributed pixel data
to the child. We had a similar debate for ACSIS processing with CADC
trying to decide whether the mask that was used to do baseline
subtraction was part of the provenance.
In that case I think the decision was that it should be in the
provenance (which meant that a single observation product had parents
from multiple observations since the mask was generated from the
coadd) because changing the mask would materially change the pixel
data.
--
Tim Jenness
|