[fitsbits] once FITS always FITS

[Maren Purves]

Steve Allen wrote:
> On Tue 2007-08-21T00:07:29 -1000, Maren Purves hath writ:
>> It being "once FITS always FITS" I object to making the
>> EXTEND keyword optional.
> 
> A question for those who have been around longer than I:

I'm not sure I've been around FITS for longer than you have,
but none of these:

> Does "once FITS always FITS" mean
> 
>   Anything ever explicitly allowed can never be disallowed
> or
>   Anything ever explicitly allowed must always remain required
> or
>   Anything not explicitly disallowed must always remain allowed
> or
>   Anything ever explicitly disallowed can never be allowed
> or some other rearrangment of those notions, or what?

sound like what I'd be after.

Anything ever required cannot be disallowed
and anything ever disallowed cannot be required

is about as far as I would go.

> In RFC 4047 we have
>   no change may be made to FITS that invalidates existing files
> but I'm not sure that's what everyone thinks.

By the same token: anything that writes FITS cannot write things
that are explicitly disallowed and anything that reads FITS cannot
throw errors/exit when reading valid FITS files?

(We don't write FITS files as such, we write HDS files (NDFs)
with FITS headers)

??

Maren