[fitsbits] Celsius unit in FITS keyword comments

[Forveille thierry]

  On 03/03/2016 02:17 PM, Rob Seaman wrote:
> Well, that wasn’t the question, but certainly IAU/IVOA/FITS should encourage the use of preferred units. The issue of what units to prefer is quite involved and we might bring ISO and BIPM into the fray at this point.
>
> But if a pressure sensor in a telescope dome records pounds/square inch, somebody might (hopefully or not) record that value in FITS files. Is it better to record what your measuring tools provide? Or for FITS to require performing sometimes quite involved unit conversions to meet narrowly specified standards that leave the resulting numbers unrecognizably distant from the originals?
>
> As the IAU link implied, degrees Celsius are frequently used by environmental thermometers. Focus models, etc, are likely calibrated in terms of deg C, at least for ground-based facilities. A strong case can be made that the people climbing over the telescope to mount sensors have a larger stake in specifying these units than do we. And for that matter they’ve already modified their choice of units by sliding switches to avoid Fahrenheit ;-)
>
> The most basic goal of archiving is to preserve and document legacy data. If those data are calibrated in Celsius, there needs to be some way to specify degrees Celsius. Kelvin may be preferred, but can it be required? If not, how should Celsius or any other non-preferred temperature scale be denoted?
The FITS standards explicitly states, from day 1 if my memory is 
correct,  that the only
supported units are SI, plus degrees for angles. The answer is thus, 
without any ifs or buts,
that conversion to Kelvins (and Pascals for the pressure sensor) is 
required.