Astronomical Data Formats: What we have and how we got here

by Jessica D. Mink
Smithsonian Astrophysical Observatory, 60 Garden St., Cambridge, MA 02131

Abstract:
Despite almost all being acquired as photons, astronomical data
from different instruments and at different stages in its life
may exist in different formats to serve different purposes.
Beyond the data itself, descriptive information is associated with
it as metadata, either included in the data format or in a larger
multi-format data structure.  Those formats may be used for the
acquisition, processing, exchange, and archiving of data. It has
been useful to use similar formats, or even a single standard to
ease interaction with data in its various stages using familiar tools.
Knowledge of the evolution and advantages of present standards is
useful before we discuss the future of how astronomical data is formatted.
The evolution of the use of world coordinates in FITS is presented as
an example.

1. Where We Are
The astronomical community's most widespread data format, FITS [1] is
35 years old, and interest in developing a new and improved standards
for formatting the larger and more varied types of astronomical data
being produced by more and more complicated instruments on larger and
larger telescopes is spreading [2] and [3]. Several existing options
are being proposed: HDF5, a Hierarchical Data System [4], and JPEG2000,
a widely-used image format [5], among others. The problems we face are
not all new, and I would like to cover some history about how we got
where we are and how our present solutions developed.

2. Formatting Data Through Its Life-cycle

2.1. Genesis: Origins of Astronomical Data
In the beginning, there is light. Most astronomical observations are of
photons. In addition to a count or other measure of their intensity, we
record information including direction of the source, wavelength,
polarization, or frequency or energy of individual photons or some
grouping thereof, and the time(s) at which they were collected.
Associated metadata describing the conditions under which the data was
created may be included with the data as a header, trailer, or internal
labels, or may reside in a separate format, such as a logbook, digital
table, or label on the data container.

At its simplest, a data format includes data structured in some way to
make it retrievable. It may be a qualitative drawing in a logbook, such
as Galileo's drawing of Jupiter and its largest satellites, with
descriptive information about the data written right next to it. It may
be a table of numbers in a published paper or monograph, with the text
of the paper providing the contextual metadata and the headings on the
table labeling the actual numbers.

In the nineteenth century, it became possible to record a signal from
photons from the sky more directly on glass photographic plates, such as
those in the Harvard Plate Collection [6]. It is made up of photographic
plates containing images of the sky, with metadata as notes in logbooks
and on the paper jackets in which the plates are stored. Metadata for
each plate includes the pointing direction, the time and exposure of the
observation, the name of the object being observed, and who observed it.
The logbook and jacket indicate what telescope was used and where it was
located (See Figure 1). Sky coverage comes from the telescope focal plane
plate scale and the physical size of the plate. Not all useful parameters
were (or needed to be) written out because the humans using the data knew
them, so additional work has been needed to make the plate images
scientifically usable [7].

Digital data acquisition formats are usually limited to one instrument
or a class of instruments. Metadata is usually recorded digitally either
in the same files or files associated in some way within a data structure
so that processing software can learn as much as possible from its digital
input. For spacecraft observations, this would be digital telemetry; for
optical telescopes, this is likely to be some sort of image files. Both
might have associated input files, such as pointing catalogs or fiber
positions.

2.2. Transfer and Exchange
As we process that data, we end up with derived data which can take many
different forms. To exchange that data, process it with standard software,
and archive it so scientists can read it in the near or far future, the
metadata has to be standardized and well-documented. It is helpful if that
metadata travels easily with the actual bits of data.

2.3. Processing and Analysis
But the data we have acquired is not ready to be used for science. As
David Hogg noted [8],

"The data is like a noisy hash of the things we care about."

It has to be processed from raw data, including observations
and calibrations, to something that can be analyzed. Before we can look at
the data, we want it formatted in a specific way which makes it accessible
to the tools we wish to use.  In the course of processing, metadata as well
as data is changed, with information about the processing or results of
the analysis being added. It is useful if the metadata both utilizes
standard definitions and is clearly associated with the data as part of
the same file or data structure.

2.4. Archiving Data
Final disposition of data can take several paths. It can be destroyed
because it is seen to have no value or there is no space to keep it. It
can be stored on media which become unreadable, such as no-longer-readable
or slowly-decaying tape or disk formats). It can be presented in a
publication, copies of which are preserved in multiple places, though
only that part of the data content relevant to the publication will be
preserved. And finally, data may be preserved in a format which is both
persistent in content and readability.

Most ground-based and much space-based data from the 1970's through the
1990's was recorded on magnetic tapes or disks which are now next to
impossible to read. More recent data on CDROM's and DVD's will be lost
as the media degrade and readers become less common. While tapes I made
during the 1970's and 1980's are no longer readable, Hollerith cards of
photometry and software from my senior year in college are still human-
and scanner-readable (see Figure 2).

If they can, scientists share their data, analysis of it, and conclusions
from it in presentations and papers. In the age of hardcopy journals and
books, these methods were more persistent than any other method of
preservation, but 21st century astronomers tend to find articles through
ADS [9] or the ArXiv server [10] and read them online instead of in hard copy. As journals move toward online publication, standardized, retrievable formatting for the long term is becoming an issue here, too.

We now have the storage capacity to save most of the data that we take, as well as its derivatives. So far the most permanent format we are using is printed paper, with tables and graphs containing data. Photographic plates, which degrade a bit faster over time, with metadata in separate paper logs, have lasted over a century. But now most of our data and more and more of the papers and presentations which describe its meaning are digital. We need both persistent media and persistent formats to keep today's data accessible over time.

3. Standardizing Data Formats
As the inventors of FITS noted in their first paper [11]:

"Under the traditional system for data interchange in astronomy,
each institution exports data on magnetic tape in its own unique
internal format. Thus, a group of N "cooperating" institutions
would begin by creating N(N-l) format translation programs. Then,
whenever one of the institutions changes its internal format,
the other N-1 institutions have to change their corresponding
translation programs. For obvious reasons, this traditional
system has been very inefficient. It would be very nice if the
astronomical community could agree, instead, on a unique data
interchange format."

It may be in the form of structures, such as a single file with several
FITS extensions [12], or a file with metadata linking to data, such as
VO returned data packages [13] containing metadata separate from the
actual data.

FITS (Flexible Image Transport System) [1] was originally designed in
1979 as an exchange format, first used by radio astronomers in the AIPS
software system [14]. It enabled astronomers to share data without having
to maintain separate translation programs. At roughly the same time, the
more flexible and more complicated N-Dimensional Data Format (NDF) was
being developed in the U.K. Tim Jenness has explained the evolution of
that system [15].

The original simple FITS consisted of a human-readable ASCII header of
80-character lines (matching the width of Hollerith cards then used to
store and use computer software and data) and blocks of binary data
described by the header and system commands. Each of these contained an
integral number of 2880-byte blocks, padded with spaces at the end of
each unit. A basic set of standard metadata keywords was included with
the original FITS definition.  The use of FITS expanded beyond exchange
and archiving to recording and processing as computers got fast enough
that the time it took to read and write ASCII header information and
convert pixel information into internally-usable bits became increasingly
negligible.

This expanding use was aided by the ability to use FITS reading and writing
libraries such as FITSIO [16] and CFITSIO [17] to deal with input and
output from local software which worked on the bits or packages of tools
such as AIPS [14], IRAF [18] (which started with a propriety format and
added FITS [19]), and WCSTools [20] which perform sophisticated operations
directly on FITS files.

If we wish to display a FITS file, we can use a variety of tools: DS9 [21]
for images, TOPCAT [22] for tables, FV [23] for either images or tables,
and something like WCSTools [20] IMHEAD to check out the metadata.

4. Standardizing Metadata: World Coordinate Systems
The availability of the FITS data format standard with its standard header
format for metadata and system of registered conventions enabled astronomers
to concentrate on science and using its simple keyword = value metadata
format to transmit parameters of their data and models to users.

One movement to standardize metadata has been the inclusion of a set of
parameters linking pixels in an image to pointing directions in the sky
or in the case of spectra, specific wavelengths or energies. As more precise
relationships between image and spatial direction became necessary for
specific projects, a variety of world coordinate system (WCS) solutions
were developed. Because there was a standard way of defining parameters
within a FITS header, it was straightforward to implement these projections
in other software.

The original FITS format [1] included only a linear projection defined by
the keywords CRPIXn, the coordinate system reference pixel for axis n,
CRVALn, the coordinate system value for axis n at that reference pixel,
CDELTn for the the coordinate increment along axis n, CROTAn for the
rotation angle of coordinate system axis n (usually only defined for one
axis and assumed to be the same for both), and CTYPEn for the name of the
coordinate axis n.

For the 1983 IRAS satellite [24], a set of standard projections-GNOMONIC
for sky sections, AITOFF for all-sky images, and SINUSOIDAL for the
galactic plane-were used in distributed images, with parameters and comment
lines including Fortran code defining the projection in the image headers.
Table 1 shows how an AITOFF projection is presented in a FITS header.

Table 1: Aitoff all-sky projection information for converting between
image coordinates PIXEL,LINE and sky coordinates XLON,XLAT from the header
of an IRAS image.

COMMENT PROJECTION FORMULAE:
COMMENT FORWARD FORMULA; XLON0 IS THE CENTER LONGITUDE OF THE
COMMENT MAP. ARC-SINE AND ARC-COSINE FUNCTIONS ARE REQUIRED.
COMMENT R2D = 45. / ATAN(1.)
COMMENT PIX = 2.
COMMENT RHO = ACOS( COS(XLAT) * COS((XLON-XLON0)/2.) )
COMMENT THETA = ASIN( COS(XLAT) * SIN((XLON-XLON0)/2.) / SIN(RHO))
COMMENT F = 2. * PIX * R2D * SIN(RHO/2.)
COMMENT SAMPLE = -2. * F * SIN(THETA)
COMMENT XLINE = -F * COS(THETA)
COMMENT IF(XLAT .LT. 0.) XLINE = -XLINE
COMMENT
COMMENT REVERSE FORMULA; XLON0 IS THE CENTER LONGITUDE OF THE MAP.
COMMENT ARC-SINE AND ARC-COSINE FUNCTIONS NEEDED.
COMMENT R2D = 45. / ATAN(1.)
COMMENT PIX = 2.
COMMENT Y = -XLINE / (PIX * 2. * R2D)
COMMENT X = -SAMPLE / (PIX * 2. * R2D)
COMMENT A = SQRT(4.-X*X-4.*Y*Y)
COMMENT XLAT = R2D * ASIN(A*Y)
COMMENT XLON = XLON0 + 2. * R2D * ASIN(A*X/(2.*COS(XLAT)))
COMMENT
COMMENT REFERENCES:
COMMENT IRAS SDAS SOFTWARE INTERFACE SPECIFICATION(SIS) 623-94/NO. SF05
COMMENT ASTRON. ASTROPHYS. SUPPL. SER. 44,(1981) 363-370 (RE:FITS)
COMMENT RECONCILIATION OF FITS PARMS W/ SIS SF05 PARMS:
COMMENT NAXIS1 = (ES - SS + 1); NAXIS2 = (EL - SL + 1);
COMMENT CRPIX1 = (1 - SS); CRPIX2 = (1 - SL)


In the AIPS system, a set of "Non-Linear Coordinate Systems" [25]-TAN
(tangent plane), SIN (sinusoidal, typically along the galactic equator),
ARC, and NCP (centered on the north celestial pole)-were developed. By
1986 [26], four more were added: GLS (Sanson-Flamsteed sinusoidal), MER
(Mercator), AIT (Hammer-Aitoff equal area all-sky), STG (Stereographic
or zenithal orthomorphic).

In 1993 and 1994, the Space Telescope Science Institute released the
"Digitized Sky Survey", which included astrometric solutions [27] in FITS
images extracted from compressed scans of Schmidt plates from Palomar and
ESO.  This was the first attempt to provide an astometrically accurate
WCS for astronomical images, a problem which has become more acute over time.

After much discussion in the community, Marc Calabretta and Eric Greisen
[28] presented a more completely worked out set of 25 world coordinate
system transformations, along with a code library which implemented them
[29]. Figure 3 shows the how well such a tangent plane projection can be
fit matching stars in an image to a catalog using WCSTools.

But there were still problems exactly matching telescope images, and
several methods of adding information to FITS headers to refine their
astrometry were proposed. Calabretta, Valdes, and Greisen started in 2003
[30] with an expansion on the previous FITS WCS papers, but a resulting
standard has yet to be published.

In the mean time, the Spitzer space observatory needed to release data
with improved WCS information and developed the SIP convention [31],
while NOAO established the IRAF-understandable polynomial expansions,
TNX [32] (based on FITS-WCS TAN) and ZPX [33] (based on FITS-WCS ZPN).
These registered conventions are documented with the FITS standard at
the FITS Support Office at NASA/Goddard Space Flight Center [34]. In
France, the Terapix project also needed to manage distortion, and Emanuel
Bertin wrote SCAMP [35] to produce a distortion correction to an image's
WCS. WCSTools includes subroutines to decode all of these distortion
methods (see Table 2), and the AST library [36], IRAF [18], and
Astrometry.net [37] implement some of them.

Table 2: WCS projections supported by WCSTools

Code Projection
PIX Pixel WCS
LIN Linear projection
AZP Zenithal/Azimuthal Perspective
SZP Zenithal/Azimuthal Perspective
TAN Gnomonic = Tangent Plane
SIN Orthographic/synthesis
STG Stereographic
ARC Zenithal/azimuthal equidistant
ZPN Zenithal/azimuthal Polynomial
ZEA Zenithal/azimuthal Equal Area
AIR Airy
CYP CYlindrical Perspective
CAR Cartesian
MER Mercator
CEA Cylindrical Equal Area
COP Conic Perspective
COD Conic equidistant
COE Conic Equal area
COO Conic Orthomorphic
BON Bonne
PCO Polyconic
SFL Sanson-Flamsteed (Global Sinusoidal)
PAR Parabolic
AIT Hammer-Aitoff
MOL Mollweide
CSC COBE quadrilateralized Spherical Cube
QSC Quadrilateralized Spherical Cube
TSC Tangential Spherical Cube
NCP Special case of SIN from AIPS
GLS Same as SFL from AIPS
DSS Digitized Sky Survey plate solution
PLT Plate solution (SAO corrections)
TNX Tangent Plane (NOAO corrections)
ZPX Zenithal Polynomial (NOAO corrections)
TPV Tangent Plane (SCAMP corrections)
TAN-SIP Tangent Plane (Spitzer corrections)

5. Into the Future
Most of the observational astronomical community uses the FITS format
for at least some stage in the life of their data. FITS meets their needs
by being a well-defined format with embedded human-readable metadata, and
having multiple software packages capable of reading it, published
specifications, and an evolving, well-defined, published set of metadata
models and keywords. But it is not everything to everybody.  Where do we
go from here, and how do we keep the advantages of a format which has
reached across disciplines and uses over decades? As Brian Schmidt has
said [38],

" Getting standards for data in place that work requires a
consensus dictatorship. It requires collaborations between librarians,
and computer scientists to figure out how to create and maintain data
hierarchies."

As we move forward, we should be careful not to lose what we have.

6. Acknowledgements
Thanks to the Harvard Plate Collection for getting me involved in its
digitization and providing useful images and to Bob Mann and Tim Jenness
for comments which were very useful in developing this paper from a talk
at the 2014 Astronomical Data Analysis Software and Systems conference
which is summarized in [39].

References

[1] D. C. Wells, E. W. Greisen, R. H. Harten, FITS - a Flexible Image
Transport System, A&AS 44 (1981) 363.

[2] B. Thomas, T. Jenness, F. Economou, P. Greenfield, P. Hirst, D. S. Berry,
E. M. Bray, N. Gray, D. Muna, J. Turner, M. de Val-Borro,
J. Santander-Vela, D. Shupe, J. Good, G. B. Berriman, Significant Problems in
FITS Limit Its Use in Modern Astronomical Research, in:
N. Manset, P. Forshay (Eds.), Astronomical Data Analysis Software and Systems
XXIII, Vol. 485 of Astronomical Society of the Pacific
Conference Series, 2014, p. 351. arXiv:1502.05958.

[3] B. Thomas, T. Jenness, F. Economou, P. Greenfield, P. Hirst, D. S. Berry,
E. Bray, N. Gray, D.Muna, J. Turner,M. de Val-Borro, J. Santander-
Vela, D. Shupe, J. Good, G. B. Berriman, S. Kitaeff, J. Fay, O. Laurino, A.
Alexov,W. Landry, J. Masters, A. Brazier, R. Schaaf, K. Edwards,
R. O. Redman, T. R. Marsh, O. Streicher, P. Norris, S. Pascual, M. Davie, M.
Droettboom, T. Robitaille, R. Campana, A. Hagen, P. Hartogh,
D. Klaes, M. W. Craig, D. Homeier, Learning from FITS: Limitations in use in
modern astronomical research, Astron. Comput. in press
arXiv:1502.00996, doi:10.1016/j.ascom.2015.01.009.

[4] T. Jenness, Reimplementing the hierarchical data system using hdf5 (2015).
doi:10.1016/j.ascom.2015.02.003.

[5] V. V. Kitaeff, A. Cannon, A. Wicenec, D. Taubman, Astronomical Imagery:
Considerations For a Contemporary Approach with JPEG2000,
Astron. Comput. in pressarXiv:1403.2801, doi:10.1016/j.ascom.2014.06.002.

[6] J. Grindlay, S. Tang, R. Simcoe, S. Laycock, E. Los, D. Mink, A. Doane, G.
Champine, DASCH to Measure (and preserve) the Harvard
Plates: Opening the 100-year Time Domain Astronomy Window, in: W. Osborn, L.
Robbins (Eds.), Preserving Astronomy's Photographic
Legacy: Current State and the Future of North American Astronomical Plates,
Vol. 410 of Astronomical Society of the Pacific Conference
Series, 2009, p. 101.

[7] S. Tang, J. Grindlay, E. Los, M. Servillat, Improved Photometry for the
DASCH Pipeline, PASP 125 (2013) 857-865. doi:10.1086/671760.

[8] D. Bard, D. Hogg, Big everything: the future of astronomical data (Nov.
2013).
URL http://kipac.stanford.edu/kipac/big-everything-future-astronomical-data

[9] M. J. Kurtz, G. Eichhorn, A. Accomazzi, C. S. Grant, S. S. Murray, J. M.
Watson, The NASA Astrophysics Data System: Overview, A&AS
143 (2000) 41-59. arXiv:astro-ph/0002104, doi:10.1051/aas:2000170.

[10] T. Vence, One Million Preprints and Counting: A conversation with arXiv
founder Paul Ginsparg (Dec. 2014).
URL
http://www.the-scientist.com/?articles.view/articleNo/41677/title/Q-A--One-Million-Preprints-and-Counting/

[11] E. W. Greisen, D. C. Wells, R. H. Harten, The FITS Tape Formats -
Flexible Image Transport Systems, in: D. A. Elliott (Ed.), Conference on
Applications of Digital Image Processing to Astronomy, Vol. 264 of Society of
Photo-Optical Instrumentation Engineers (SPIE) Conference
Series, 1980, p. 298.

[12] J. D. Ponz, R. W. Thompson, J. R. Munoz, The FITS image extension, A&AS
105 (1994) 53-55.

[13] M. Dolensky, D. Tody, The Simple Spectral Access protocol, in: P. J.
Quinn, A. Bridger (Eds.), Optimizing Scientific Return for Astronomy
through Information Technologies, Vol. 5493 of Society of Photo-Optical
Instrumentation Engineers (SPIE) Conference Series, 2004, pp.
262-268. doi:10.1117/12.553167.

[14] I. Associated Universities, AIPS: Astronomical Image Processing System,
Astrophysics Source Code Library ascl:9911.003 (Nov. 1999).

[15] T. Jenness, D. Berry, M. Currie, D. P.W., E. F., N. Gray, B. McIlwrath,
K. Shortridge, M. Taylor, P.Wallace, R.Warren-Smith, Learning from
25 years of the extensible n-dimensional data format (2015).
doi:10.1016/j.ascom.2014.11.01.

[16] W. D. Pence, FITSIO - A New Fortran-77 Subroutine Interface for Reading
and Writing FITS Format Files, in: Bulletin of the American
Astronomical Society, Vol. 23 of Bulletin of the American Astronomical
Society, 1991, p. 936.

[17] W. Pence, CFITSIO, v2.0: A New Full-Featured Data Interface, in: D. M.
Mehringer, R. L. Plante, D. A. Roberts (Eds.), Astronomical Data
Analysis Software and Systems VIII, Vol. 172 of Astronomical Society of the
Pacific Conference Series, 1999, p. 487.

[18] National Optical Astronomy Observatories, IRAF: Image Reduction and
Analysis Facility, Astrophysics Source Code Library ascl:9911.002
(Nov. 1999).

[19] N. Zarate, P. Greenfield, A FITS Image Extension Kernel for IRAF, in: G.
H. Jacoby, J. Barnes (Eds.), Astronomical Data Analysis Software
and Systems V, Vol. 101 of Astronomical Society of the Pacific Conference
Series, 1996, p. 331.

[20] J. D. Mink, WCSTools: Image Astrometry Toolkit, Astrophysics Source Code
Library ascl:1109.015 (Sep. 2011).

[21] Smithsonian Astrophysical Observatory, SAOImage DS9: A utility for
displaying astronomical images in the X11 window environment,
Astrophysics Source Code Library ascl:0003.002 (Mar. 2000).

[22] M. Taylor, TOPCAT: Tool for OPerations on Catalogues And Tables,
Astrophysics Source Code Library ascl:1101.010 (Jan. 2011).

[23] W. Pence, P. Chai, Fv: Interactive FITS file editor, Astrophysics Source
Code Library ascl:1205.005 (May 2012).

[24] C. A. Beichman, G. Neugebauer, H. J. Habing, P. E. Clegg, T. J. Chester
(Eds.), Infrared astronomical satellite (IRAS) catalogs and atlases.
Volume 1: Explanatory supplement, Vol. 1, 1988.

[25] E. W. Greisen, Non-linear Coordinate Systems in AIPS (Jun. 1983).
URL ftp://ftp.aoc.nrao.edu/pub/software/aips/TEXT/PUBL/AIPSMEMO27.PS

[26] E. W. Greisen, Additional Non-linear Coordinate Systems in AIPS (Jan.
1993).
URL ftp://ftp.aoc.nrao.edu/pub/software/aips/TEXT/PUBL/AIPSMEMO46.PS

[27] J. L. Russell, B. M. Lasker, B. J. McLean, C. R. Sturch, H. Jenkner, The
Guide Star Catalog. II - Photometric and astrometric models and
solutions, AJ 99 (1990) 2059-2081. doi:10.1086/115484.

[28] E. W. Greisen, M. R. Calabretta, Representations of world coordinates in
FITS, A&A 395 (2002) 1061-1075. doi:10.1051/0004-6361:20021326.

[29] M. R. Calabretta, Wcslib and Pgsbox, Astrophysics Source Code Library
ascl:1108.003 (Aug. 2011).

[30] M. R. Calabretta, F. Valdes, E. W. Greisen, S. L. Allen, Representations
of distortions in FITS world coordinate systems, in: F. Ochsenbein,
M. G. Allen, D. Egret (Eds.), Astronomical Data Analysis Software and Systems
(ADASS) XIII, Vol. 314 of Astronomical Society of the
Pacific Conference Series, 2004, p. 551.

[31] D. L. Shupe, M. Moshir, J. Li, D. Makovoz, R. Narron, R. N. Hook, The SIP
Convention for Representing Distortion in FITS Image Headers,
in: P. Shopbell, M. Britton, R. Ebert (Eds.), Astronomical Data Analysis
Software and Systems XIV, Vol. 347 of Astronomical Society of the
Pacific Conference Series, 2005, p. 491.

[32] D. Tody, L. Davis, F. Valdes, TNX convention for Representing FITS Image
Distortions (2008).
URL http://fits.gsfc.nasa.gov/registry/tnx.html

[33] F. Valdes, ZPX convention for Representing FITS Image Distortions (2011).
URL http://fits.gsfc.nasa.gov/registry/zpxwcs.html

[34] W. D. Pence, The FITS Support Office at NASA/GSFC (2014).
URL http://fits.gsfc.nasa.gov

[35] E. Bertin, SCAMP: Automatic Astrometric and Photometric Calibration,
Astrophysics Source Code Library ascl:1010.063 (Oct. 2010).

[36] D. S. Berry, R. F. Warren-Smith, AST: World Coordinate Systems in
Astronomy, Astrophysics Source Code Library ascl:1404.016 (Apr.
2014).

[37] D. Lang, D. W. Hogg, K. Mierle, M. Blanton, S. Roweis, Astrometry.net:
Astrometric calibration of images, Astrophysics Source Code
Library ascl:1208.001 (Aug. 2012).

[38] Astronomical Data and Astronomical Digital Stewardship: An interview with
Brian Schmidt by Jane Mandelbaum (Nov. 2013).
URL
http://blogs.loc.gov/digitalpreservation/2013/11/astronomical-data-and-astronomical-digital-stewardship-an-interview-

[39] J. Mink, R. G. Mann, R. Hanisch, A. Rots, R. Seaman, T. Jenness, B.
Thomas, W. O'Mullane, The Past, Present and Future of Astronomical
Data Formats, ArXiv e-printsarXiv:1411.0996.