file naming

[<aah@nofs.navy.mil>]

Tom recently suggested that I take an active part
in any file naming discussion, since I wrote a TN
on the process a while back.  That technical note
was concerning the Mark III data, a much more limited
project.
  My feelings towards the Mark IV have already been
stated:
  (a) files stored on CDROM should be in an 8x3 format
      so that they obey the ISO 9660 standard.
  (b) if you need more characters than that to uniquely
      identify the data, then you should use a directory
      structure on the CDROM.

I've never been a great fan of including the 'time' as part
of a file name; the number of characters necessary to obtain
a unique name depends on the exposure length.  I have always
been a great fan of a sequence number, so that automated
processing tasks can easily access the files.
  The ways most professional observatories name files
usually falls into two categories:
  (1) a filename based on the name of the observed field,
      such as ngc4449.fit.  This can get long if you
      start including filters, sequence numbers, on/off sky, etc.
  (2) a filename based on the UT date plus a sequence
      number.  Examples are y01d138.005, 990504.101,
      j1520f233.fit (where the last one is based on the
      last 4 digits of the Julian Date plus a 3-digit file number).

So if I were to state a preference at this stage, I would
suggest the following:
  (1) use a directory structure on the disk, where the
      upper level directory indicates the site.  How the
      site is to be named is TBD; perhaps by the system
      'name' like TOM, MIKE, etc.      
  (2) under the site directory, use a unique filename with
      a sequential file number.  I would suggest
         jddddxsss.fit
         j  -- indicates following digits are Julian Date
         dddd - last 4 digits Julian date (9999 days is longer than
               these systems will run)
         x  -- single character of filter name (bvri)
         sss - 3 digits sequence number, starting at 000
           (1000 files should be plenty for a Mark IV)

My personal data is stored in a similar manner, though the
file names look more like
    yymmddfx.sss
     yy - (year - 2000)
     mm - month
     dd - day
     f  - filter name
     x  - processing step (r=raw, d=darksub f=darksub&flatfield)
     sss - sequence number
but I understand that this does not uniquely identify the files
as 'fits' and some processing programs can't handle it.  I prefer
my format since I think in yymmdd instead of jdddd.
Arne