[fitswcs] DCF proposals

[Mark Calabretta]

On Tue 2001/10/16 09:38:20 MST, Frank Valdes wrote
in a message to: fitswcs at NRAO.EDU

Dear Frank,

Looks to me like there are no major disagreements on this.  Particular
points of contention are

* Maximum index of the PVj_ms, DVj_ms, etc.: the repercussions for memory
  usage by WCS header parsers is real and the memory usage figures in the
  DCF notes are not worse case. 

  Systems without dynamic memory allocation will have to reserve static
  memory for the full range of indices at compile time.

  Systems with dynamic memory allocation will probably want to allocate
  memory for the full range of indices when the first DVj_ms is
  encountered in the header.

  Multiplying factors are the maximum number of coordinate axes and the
  maximum number of simultaneous WCS to be handled.  Consequently the
  memory requirements are not insignificant, even given the postmodern
  assumption of unlimited memory which, in any case, can't be assumed for
  Joe Bloggs wanting to read a FITS file on his home PC.

  Therefore it is prudent to restrain the maximum index.

* Index sparseness: it's true that the density of indices of non-zero
  parameters is irrelevant for keywords in the primary header.

  However, as discussed in point 5 of the email notes, as there will be
  at least 1000 DVj_ms and FITS tables are limited to 999 columns, we
  will need a way to store all of the jDVn_m in one table column as an
  array.  The presence of a large number of zero-valued parameters in
  such an array will bloat the table, so clearly a tradeoff must be
  made between index efficiency and index density.

  I had thought of retaining both polynomial representations but applied
  occam's razor.  However, since the first form can always be translated
  into the second (but not vice versa), and it is simpler and more
  efficient in terms of the number of keywords, I agree that it would be
  worth having both representations for use as appropriate.  Maybe others
  will disagree.

* PSj_ms, DSj_ms: to be discussed.

There's probably not much point going further with this discussion until
the first draft of Paper IV is ready.

Cheers, Mark