On Wed, 15 Jul 2009, Mark Taylor wrote:
> On Wed, 15 Jul 2009, Peter W. Draper wrote:
>
> > On Tue, 14 Jul 2009, Tim Jenness wrote:
> >
> > > On Wed, 15 Jul 2009, Peter W. Draper wrote:
> > >
> > > > I also suspect that the elephant in the room against 1D FITS is
> > > > handling the WCS information. That's considerably simplified using
> > > > a table.
> > > >
> > >
> > > but AST handles it no problem :-;
> > >
> > > So it's been decided that everything is a LutMap?
> >
> > Yes. For now anyway. No doubt given time this will get back to where we
> > were (requires STC they'll say).
>
> Maybe. But unlike for 2+ dimensional arrays I don't really see what's
> so great about using WCS rather than a LUT for 1d. And if you can avoid
> reading the FITS paper and writing the software to make sense of it,
> that's good news (AST's "no problem" presumably covers several years of
> David's blood sweat and tears). Using WCS presumably saves you a factor
> of ~2 in file size - does it have other major benefits?
The main benefit was following a well thought out existing standard which
has most of the things in the WCS area already sorted. The problem, which
I'm happy to agree with, is that the data model is quite restrictive, so
something proabably had to give. Fine, but 90% of all spectra could still
be covered by the 1D FITS standard, so the jumping ship by the VO world
seems convenient for them and unnecessary work to the rest (although I
know some domains had already never really adopted 1D, so there are also
many table spectra out there as well).
None of this touches the additional consideration that we need the
spectral WCS anyway to complete datasets like spectral cubes.
Peter.
|