Re: FW: Data to be collected in the 'master?' database

[<aah@nofs.navy.mil>]

Tom suggests adding version number to the fits header.
I think that the acquisition program name and version
number needs to be part of the fits header for sure.
Any subsequent modification to the fits file (such as
flatfielding) then needs to add a comment line to the fits
header to indicate what program made the modification and,
in general, what modification was performed.
  However, for the pipeline, no modification to the fits
file is made.  The program name and version number need
to be added to the starlist header, not to the fits file.
  I think Tom was suggesting (below) that the source code for each
program version be archived on the web site, again as a development
tracer so that any data in the master database has a fully
documented history.  This is a good suggestion though it
might be difficult to implement.  At least a revisions note
file needs to be included on the web site.
Arne
-------------
I second Arne's suggestion here.  As you all may have noticed, I always 
write the data taking program on the data disks.  Thus we have the whole 
program, not just some version number that might not have been changed 
after some fix.  The data sets are so large that the program takes up 
insignificant space.  This is probably not practical for reference catalogs 
though.

Seems to me that the fits header is the right place for this.  A processing 
step should write the version numbers of the programs used into the 
header.  This then requires that we stash away each version number 
somewhere.  I think the web site is the right place.  Then with normal back 
ups to CD ROM we will always be able to find it.  This assumes we will be 
able to read CD ROMs forever.  We will not.  Sigh!

Tom Droege