[fitsbits] MIME for compressed FITS?

[Steve Allen]

On Tue 2007-05-22T15:21:13 -0700, Rob Seaman hath writ:
> An application/FITS can be any conforming FITS object such as MEF -
> but few clients will know what to do with it.

This was the problem that finally overcame the energy barrier to promp
the creation of the FITS MIME document.  Bill Joye was trying to
figure out how ds9 could tell just what it was supposed to do with a
purportedly FITS file that it obtained by any means, but in particular
he was hoping that MIME types might provide a solution.  By the end of
the FITS MIME effort it became clear that the solution of this problem
is not possible within the Internet mechanisms but is internal to
FITS.

> What is a FITS tile compressed SIF file?  It remains an application/
> FITS, but can it be promoted (if that is the right word) to image/
> FITS with a different content encoding?  (It remains a coherent
> representation of a single image.)  Steve appears to assert "no" and
> that the vast internet community will have no truck with the (even
> vaster in physical domain) astronomical community.

Given the last year's tweak in the tile compression mechanisms a FITS
tile compressed SIF file (which must consist of a PHDU plus an
extension HDU that is a table and is thus application/fits) can now be
isomorphic with a classical FITS file with a single PHDU (which can
be image/fits).

content-(en)coding is not directly relevant to this.

> The alternative is to turn application/FITS into something useful, I
> suppose.

That's it.  FITS needs mechanisms for saying
    This file (and/or HDU) makes use of the following list of features
    and conventions.
With that in place a FITS application might ascertain whether it has
any interest or ability to make sense of the content which has been
stuffed into the zoo of extension HDUs.

> Do we need to consider another round of MIME murmurings?  Failing
> that, is MIME really of any value to FITS?

MIME is essential to those VO exchanges of data which rely on HTTP as
a transfer mechanism.

This sort of problem is not specific to the FITS community.  At the
time of the FITS MIME effort, however, not even the internet community
had a standardized mechanism for describing the transmission of a file
of given MIME media type and compression via the mechanisms described
for SMTP.  The specification for HTTP has the concept of
content-(en)coding, but the specification for sending MIME media types
in e-mail messages did not.

Anyhow, Bill Joye would love to see this problem addressed.
As things stand ds9 basically has to be told which set of heuristics
to use when it is given a FITS file, and the interpretation of all
other complex data stored in FITS is equally reliant on implicit
agreements between the creator and consumer.  The IAUFWG effort to
make a list of recognized FITS conventions is one step toward the
solution.

--
Steve Allen                 <sla at ucolick.org>                WGS-84 (GPS)
UCO/Lick Observatory        Natural Sciences II, Room 165    Lat  +36.99855
University of California    Voice: +1 831 459 3046           Lng -122.06015
Santa Cruz, CA 95064        http://www.ucolick.org/~sla/     Hgt +250 m