[fitsbits] further reopening of Public Comment Period on the CONTINUE convention

[Mark Calabretta]

On Wed, 20 Apr 2016 15:08:58 -0700
Steve Allen <sla at ucolick.org> wrote:

I can feel a meta-discussion coming on...

> When the WCS conventions were added to FITS it was not required that
> all FITS readers be upgraded in order to make proper interpretation of
> their full meaning,

Historically, the reason unanimity was required for accepting the WCS
proposal was that there was an expectation that existing software would
(eventually) be upgraded to understand it.

>and it was not required to rewrite existing FITS
> files which happened to have keywords matching the name in the new
> standard.

The WCS proposal was careful not to invalidate existing FITS files that
used the AIPS convention.  Indeed, Paper II has a section on how they
should be handled.
 
> Old FITS readers could still read new files with WCS keywords without
> syntax error, but not get the meaning.

Worse, they might get the wrong meaning - I know of at least one WCS
interpreter that produces incorrect coordinates.  Like I said, it was
expected that WCS interpreters would be updated, and library software
(WCSLIB, etc.) was provided to assist with that.

> New FITS readers could still read old files with WCS keyword names
> whose values were not conformed to the new requirements, and in that
> case those new FITS readers would have to gracefully fail.

As above, the AIPS convention was treated as a de facto standard, and
Paper II provides translations.  WCS interpreters are expected to apply
them and gracefully produce correct answers.
 
> I am not seeing how the CONTINUE convention for string-valued keywords
> is significantly different from previous changes to FITS conventions.

The proposal to adopt CONTINUE, as currently presented, has a cost in
terms of the amount of existing software that would need to be modified
to account for the possibility, however unlikely, that the existing
string-valued, reserved keywords are allowed to be CONTINUE'd.

Is this COST justified by the touted BENEFITS?

I voted NO!

But there is a SIMPLE (= T) alternative.  Don't allow existing
string-valued, reserved keywords to be CONTINUE'd.

> It still looks like the FITS community, readers and writers, can
> coexist by following RFC Editor Jon Postel (RIP) advising all parties
> "Be conservative in what you send, be liberal in what you accept"

In this case, being "liberal in what you accept" would require
changing WCSLIB's API (and ABI), which, as I have said, I am not going
to do.

Regards,
Mark Calabretta