[fitsbits] Infinite floats in FITS headers

[Peter Weilbacher]

On Mon, 5 May 2014, Lucio Chiappetti wrote:

> On Mon, 5 May 2014, Peter Weilbacher wrote:
>
> > Quick question: what should I do when I have an infinite floating point
> > value (like INF, NAN) that I want to save into a FITS header?
> > I do not explicitly) find this case in the FITS specifications.
>
> It is not handled by the spec. It was discussed at least twice (you can
> look in the fitsbits archives) in a thread in August 1997 and in a more
> recent one in June 2013. The latter was discussing pro and cons and
> suggesting some hacks.

Lucio, thanks for pointing me to those discussions, it appears that I
was unable to properly search...

So I guess this matter has not evolved in any way since then. I found it
strange that in the discussions IEEE infinite values were regarded as
problematic because apparently some platformat are not supporting them.
But since for FITS these values are already allowed in the data, this
reservation is probably not valid.

Best regards,
   Peter.
-- 
Dr. Peter Weilbacher                http://www.aip.de/People/PWeilbacher
Phone +49 331 74 99-667                       encryption key ID 7D6B4AA0
------------------------------------------------------------------------
Leibniz-Institut für Astrophysik Potsdam (AIP)
An der Sternwarte 16, D-14482 Potsdam

Vorstand: Prof. Dr. Matthias Steinmetz, Dr. Ulrich Müller
Stiftung bürgerlichen Rechts, Stiftungsverz. Brandenburg: 26 742-00/7026