[fitsbits] start of Public Comment Period on the Green Bank convention

[Lucio Chiappetti]

Another late reply for the records

On Tue, 23 Jun 2015, Mark Calabretta wrote:

> The Green Bank convention applies to most keywords in tables, not just 
> the WCS keywords, so belongs in Chapter 7, not 8.  Perhaps make a new 
> section, 7.4.

Actually that was a suggestion I raised. But the convention was born in 
the WCS domain, so at the end it was left in chapter 8.

If we make a 7.4 for TLMIN/TLMAX (7.4.1) we can make a 7.4.2 for Green 
Bank.

> Clearly, there are many keywords that the Green Bank convention cannot
> apply to: EXTNAME, EXTVER, EXTLEVEL, XTENSION, BITPIX, NAXIS, NAXIS1,
> NAXIS2, PCOUNT, GCOUNT, TFIELDS, EXTVER, TFORMn, TTYPEn, TUNIT, TDIMn,
> THEAP, and probably others.  As a standards document, they all have to
> be identified and listed.

You refer, I suppose, to the last sentence of the green parts of pag. 26 
and 29. There was an older version mentioning "mandatory or reserved" for 
which I raised an objection similar to yours.  The reference to "mandatory 
keywords" was removed (of course Green Bank does not make sense for them, 
and most of what you list are mandatory kwds).

So I interpreted in my mind "reserved" as "reserved but not mandatory",
and was satisfied (wrongly ?).

An explicit exclude list (which may include in summary "all mandatory 
kwds" and of course TTTPEn to avoid an absurd recursion) is obviously 
better.

> The statement on p29
>
>  It is *strongly recommended* that if the value of a TTYPEn keyword is
>  the same as the name of a reserved keyword, than the quantity in that
>  field should conform to the definition of that keyword.
>
> should be changed to
>
>  If a TTYPEn keyvalue is the same as the name of a reserved keyword
>  not on the list of exclusions [as above], then [not "than"] the
>  quantity in that field *shall* be interpreted according to the
>  definition of that keyword.
>
> because what other sense can software make of it?  Or even a human
> reader for that matter.

My interpretation was "	the quantity should have the SAME DATA TYPE as 
the keyword". I suggested something like that during the editing (replace 
the word "quantity" with "data type").

> There should be a statement regarding precedence if a keyword and column
> of the same name are both present.

perhaps, or see below ...

> There should be some mention of the fact that columns can only be
> collapsed if the column name can become a legitimate FITS keyword.

You mean the column name is ANY legal keyword name (not necessarily a 
reserved keyword name, also a project specific one) ?

> Sect. 8.2.1 should retain the comments about TFORMn = 'CRVALia' being
> allowed for historical compatibility, though that usage should be
> deprecated.

Agreed

> FITS readers normally only need to parse keywords and read keyvalues.
> Implementing the Green Bank convention requires that they also parse
> keyvalues associated with TTYPEn.

During the discussion I proposed a description of this sort, but it was 
considered excessive for the standard document. Should it be re-fished ?

    - if a particular keyword 'kwdnam' EXPECTED because of the peculiar
      nature of the FITS file (known by EXTNAME or other signature but not
      otherwise specified by the standard[*]) is MISSING, the direct Green
      Bank convention implies the reader has to check for a TTYPn='kwdnam'.
      If it finds it, should be use it row by row. If it does not find it,
      should assume a default constant value, or fail.

      [*] except in a few cases, GTI files, compressed images or
      tables, spatial region files.

      Note that 'kwdnam' need not to be reserved by the standard. It
      could also be a keyword required by the peculiar nature of the
      FITS file.

    - the reverse Green Bank convention occurs when a particular named
      column TTYPn='colnam' is EXPECTED because of the peculiar nature of
      the FITS file (known as above) but the column with such name is
      missing [x]. Hence the reader has to check for a keyword named
      'colnam' and use such constant value for all rows, or otherwise a
      suitable constant default

      [x] what does missing mean ? NO TTYPn='colnam', or a TTYPn='colnam'
      and a TFORMn specifies zero-length ?

This will implicitly specify your "preference" (but it depends on the 
expectation for the particular file)


> This leads me to suggest that there ought to be some standard way to 
> signal that the Green Bank convention is in use.

This brings in the old argument of "convention signatures". In some cases 
the signature is implied (INHERIT, TLMIN/TLMAX, CHECKSUM, CONTINUE are 
conventions which are related to the presence of some kwd; compress 
requires ZIMAGE or ZTABLE), But actually we removed those signatures where 
present (LONGSTRN).  Introducing a signature will introduce a(n 
unpleasant) break between new files following the "standard convention' 
(if signed) and old files following the old "registered convention" 
(unsigned).


-- 
------------------------------------------------------------------------
Lucio Chiappetti - INAF/IASF - via Bassini 15 - I-20133 Milano (Italy)
For more info : http://www.iasf-milano.inaf.it/~lucio/personal.html
------------------------------------------------------------------------
Do not like Firefox >=29 ?  Get Pale Moon !  http://www.palemoon.org