Deprecation of Random Groups

[Tim Pearson]

>From the proposed standard (Section 7): "The binary table extension
... can accommodate the structure described by random groups."  I
think that the binary table extension can also accommodate the
structure described by the Primary Data Array, but we have not
"deprecated" that.

Software exists to read and write data in the random groups structure
(e.g., cfitsio), so why not let application designers decide for
themselves whether the random groups structure or a binary table is
more appropriate?  Putting data in a binary table format does not make
it immediately accessible - applications still have to be written that
understand the physical or astronomical meanings of the various
columns. This is presumably why the binary table format has not
replaced random groups in radio interferometry.

I would welcome more discussion of the relative merits of random
groups and binary tables. The only advantage of a binary table that I
can see is that generic software can - in some degreee - display,
edit, or restructure the table, but this does not particularly help in
designing a real application that needs to know the meanings assigned
to the rows and columns of the table.

Perhaps everybody would be happy with a slight change of wording, such
as: "The binary table extension can accommodate the structure
described by random groups, and offers advantages over the random
groups format, so it is recommended for new applications," rather than
"its [random groups format's] use for future applications has been
deprecated", with the implication (Section 3) that a "deprecated"
structure is "obsolete".

- Tim Pearson