[fitsbits] DOI keyword usage for FITS? {External}

[Seaman, Robert Lewis - (rseaman)]

Outlook is a terrible mail client. In case it wasn’t clear, I was quoting from the current FITS standard. To repeat, from section 4.1.2.3:

“The mandatory FITS keywords defined in this Standard must not appear more than once within a header. All other keywords that have a value should not appear more than once. If a keyword does appear multiple times with different values, then the value is indeterminate.”

And REFERENC= T violates section 4.4.2.3. Also, rather than define a separate table for each type of metadata, FITS could retire the entire header metadata model (beyond structural keywords) in favor of whatever normalized set of metadata bintables can best represent a complete modern data model.

Many discussions over the years, e.g., at ADASS, have been in support of building near-eternal astronomical data archives. During the early days of FITS, this was supported by its documentation being published in paper journals. I don’t know what digital journals are planning for long-term preservation, but they certainly should plan for their intellectual property to survive their own inevitable bankruptcies whenever some vulture capitalist comes along.

Note that abstract submission is open for ADASS XXXIII: https://adass2023.lpl.arizona.edu, registration will open soon. This discussion is already a good start at the agenda for the FITS BoF. Somebody should submit an abstract.

Remind me not to be buried in the Netherlands.

Rob

On 8/4/23, 1:16 AM, "fitsbits" wrote:

Why not define an extension table "REFERENCES"?
This will allow multiple references to the same underlying data
as well as references to multiple data sets, e.g. calibration tables
used to reduce the current data.

A row in the table might include the following columns:

RELEVANCE  (To what does this reference refer.)
TYPE       (DOI,  ESO Archive, VIZIER ...)
REFERENCE  (specific link)
KEYWORDS   (necessary keys to find data in an archive)

..


You might include a keyword:
REFERENC =                       T
in the main header to indicate that the table exists.

As to DOI dropping links if you don't pay the rent, this reminds me of
the
Dutch graveyard policy of disinterring the remains of deceased whose
relatives have not paid the rent and dumping them in a common bone pit.

However, it is absurd to think that any archive is eternal.  What
happens to astronomical literature links if Springer or IOPscience go
bankrupt?


Walter

On 2023-08-04 09:43, Mark Calabretta via fitsbits wrote:
> Record-valued keywords, e.g. DPj and DQi as defined in WCS Paper IV,
> are multiply defined by design.
>
> As the original FITS specification did not prohibit multiply defined
> keywords, the "once FITS always FITS" principle deems that they cannot
> subsequently be prohibited.
>
> I see no reason not to allow multiply defined keywords, provided that it
> is known in advance that a particular keyword may be multiply defined so
> that header parsing software is prepared for the possibility.
>
> Mark Calabretta
>
>
>
> On Thu, 3 Aug 2023 17:18:33 +0000
> "Seaman, Robert Lewis - \(rseaman\) via fitsbits"
> <fitsbits at listmgr.nrao.edu> wrote:
>
> That seems a bit dismissive of the whole notion of DOIs. Also,
> thoughts about duplicate keywords?
>
>
> The mandatory FITS keywords defined in this Standard must not appear
> more than once within a header. All other keywords that have a value
> should not appear more than once. If a keyword does appear multiple
> times with different values, then the value is indeterminate.
>
> Rob
>
> _______________________________________________
> fitsbits mailing list
> fitsbits at listmgr.nrao.edu
> https://listmgr.nrao.edu/mailman/listinfo/fitsbits

_______________________________________________
fitsbits mailing list
fitsbits at listmgr.nrao.edu
https://listmgr.nrao.edu/mailman/listinfo/fitsbits
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listmgr.nrao.edu/pipermail/fitsbits/attachments/20230804/d74174d2/attachment-0001.html>