[iaufwg] convention or standard

[William Pence]

I generally agree with the recent comments on this subject, and will just
add a few more comments if anyone is still interested...

forveill at cfht.hawaii.edu wrote:
> we should by all means avoid any
> misunderstanding that the conventions have any "official"
> status. They are something that somebody uses/used and it's
> good that they are documented, but given the very loose standard
> (essentially some decent description) for "accepting" a convention
> the distinction must be made very clear.

I've added the following paragraph near the top of the Registry web page:

"It should be emphasized that the registration process is mainly designed to
ensure that the documentation about the convention meets a minimum level of
completeness and clarity. These conventions are not necessarily endorsed by
the IAU FITS Working Group. A separate and more rigorous review process is
required before a FITS convention is approved as part of the FITS Standard."

Any suggestions for improving this web page are welcome.  I expect that the
layout of the information on the page will change as more conventions get
added (maybe into the form of a table).

Mark Calabretta wrote:
> I may stand to be corrected, but I think in practice that conventions
> defined by one package have not generally been taken up by others.

Having the central registry of convention may help make them more widely
reused.

> The "conventions in the standard" are no longer "conventions", they're
> now standards.  Referring to them as conventions, however qualified,
> blurs the distinction between the two.

I see your point, but I think it is common usage to refer to some of the
things in the FITS Standard as "conventions". Even the FITS Standard itself
uses this terminology in several places (e.g., the random groups convention,
and the convention for representing unsigned integers).

>> conventions in the Registry for approval into the Standard, simply because 
>> the procedure is too cumbersome, and there is no real incentive to go 
>> through that whole process.  An alternative might be to create a 2-tier set 
> 
> That's an indictment on the FITS standards process.  That process is the
> IAUFWG's responsibility.  Therefore it's an indictment on us.

I believe the IAUFWG will be responsive to whatever it thinks is important
to do so I don't think we have to worry about issuing indictments yet. :-)

Lucio Chiappetti wrote:
> I believe (as I just said on FITSBITS) that there are two more sources of 
> confusion in the present discussion :
> 
> - one is that there are presently two different "registries" involved,
>   one is the convention registry 
>   http://fits.gsfc.nasa.gov/fits_registry.html
>   the other one is the list of registered extensions
>   http://fits.gsfc.nasa.gov/xtension.html

I think this minor source of confusion can be eliminated by a suitable
reorganization of the information on FITS Support Office web site.  I'll
work on this as soon as I can, and will follow Preben's suggestion to add a
link to the list of registered extensions on the top page.

> - the other source of confusion is that the second "convention" to be
>   discussed (XTENSION='FOREIGN') is actually the definition of a new
>   conforming extension so it falls in both categories !


EXTENSION = 'FOREIGN' will be listed in both places.  I don't expect there
are many other extension types that need to be registered, but if there are
any, then they will also be entered in both places.  I think the short list
of extension names is useful and shouldn't be eliminated just because we now
have the more general registry of conventions.

> But the matter of a conforming extension is different ! It defines an 
> altogether different file organization. There is nothing I can do if I 
> receive a file with XTENSION='xxxx' and that is not one of the standard 
> ones, then choke and say "unsupported extension", or write a new reader.

I had a similar initial reaction, but after thinking about it more I don't
agree.  In principle, a general FITS reader should be able to do the
following with any conforming extension, even if it has never seen that type
of extension before:

   - read and write keywords in the header
   - read the bytes in the data unit (by default the data unit can be
     interpreted as a stream of unsigned 8-bit bytes).  It could also
     write a stream of bytes to the data unit, but it will not necessarily
     know how to update the size information in the header keywords.
   - skip over the extension to the next HDU in the file, if any.

On the other hand, I still think it is not a good idea to create new
extension types because most existing FITS software was probably not written
with this degree of generality in mind, and so will likely behave strangely
when presented with an unknown extension type.  Human users will also be
confused by any extension types.

Preben Grosbol wrote:
>  1) The (XTENSION='FOREIGN') issue should not be treated as a convention
>      but as a 'local' FITS extension like XTENSION='A3DTABLE'.  

I would argue that XTENSION='FOREIGN' is a convention, at least in the
commonly understood meaning of the word.  The only restriction the FITS
Standard places on new conforming extensions is that the name must be
registered with the IAUWFG.  The 'Registry of FITS Conventions' is the
logical place to provide a more complete description of the extension.

regards,
Bill
-- 
____________________________________________________________________
Dr. William Pence                       pence at milkyway.gsfc.nasa.gov
NASA/GSFC Code 662       HEASARC        +1-301-286-4599 (voice)
Greenbelt MD 20771                      +1-301-286-1684 (fax)