[iaufwg] FITS Registry Proposal

[Mark Calabretta]

On Tue 2006/03/21 23:00:34 CDT, William Pence wrote
in a message to: IAU-FWG <iaufwg at nrao.edu>

Dear Bill,

>convention that has been adopted by every mission in the past 25 
>years  (the HEASARC alone has 100s of instrument-specific conventions
>for specific keyword and column names but I don't think most of them have
>much potential for reuse by any other mission).  This is not intended

We agree that specific keyword and column names would not constitute a
"convention".  If you think of FITS as a grammar then a "convention"
would be a language based on the FITS grammar and a keyword or column
name would just be a word in the language.

Apart from SDFITS, another example would be a description of the set of
WCS keywords used by DSS.  It would be interesting to know just how many
conventions there are out there.  Who can offer specific examples?

>to be a huge FITS keyword data dictionary where every keyword that
>anyone ever used is documented (although this would also be useful
>if anyone has the time and energy to implement it). Instead, I see this 

Also I think best implemented as a wiki.

>as mainly for conventions that could potentially be reused for the same 
purpose
>by other missions or projects.  Certainly the SDFITS convention
>fits into this category.

Sure, but the other side of the coin is to have an authoritative 
repository of conventions so that data that has already been archived
can be interpreted at some arbitrary time in the future.

>So, the real issue here is how do we define what is meant by the
>"important and significant" conventions that are worthy of being
>put into the IAU-FWG Registry of Conventions.  The rules I proposed

If at least one person thinks it's worth the effort to add a description
of a convention to the repository, and is prepared to do it, and do it
properly then that should be sufficient.  What do you gain by
restricting contributions?  It's just another entry in the index.

Another example: the ATNF's internal format is called RPFITS which
actually is not FITS (Eric calls it Ray's Pseudo-FITS, see
http://www.atnf.csiro.au/computing/software/rpfits.html).  But RPFITS
looks a lot like FITS except that it begins with SIMPLE = 'FALSE', has
a block size of 2560 bytes, uses VAX floating point representation, and
a few other things, but otherwise looks just like FITS.  

ATNF data (ATCA, Parkes, Mopra, Tidbinbilla) has been archived in this
pseudo-FITS format for the last couple of decades.  I believe it would
be appropriate to have an entry for RPFITS in the registry simply to
inform people that, contrary to appearances, RPFITS is not FITS, to
describe briefly how it differs, and tell them where to get software
that reads it.

>were a first attempt at this definition.  Do you think that the way 
>those rule are currently worded that conventions like SDFITS would
>be excluded?   If so, can use suggest some alternate wording?

The current wording leans strongly towards reusing a convention for
future data taking rather than describing it for the purpose of
interpreting data that is already archived.  I don't want to suggest
alternate rules or wording at this stage but I believe the intent should
be as above - if at least one person thinks it's worth the effort to add
a description of a convention to the repository, and is prepared to do
it, and do it properly then that should be sufficient.

This is why I favour something based on the wikipedia model:
http://en.wikipedia.org/wiki/Wikipedia:About, probably with registered
contributors.  The main problem then is to ensure that the documentation
is complete.  In wikipedia this is done simply by adding a warning to
the top of a wiki entry stating that it is incomplete and needs work.

Regards, Mark