| Quick-Start Introduction to PDS Archiving |
 |
Data Formats
PDS imposes very few hard
restrictions on the physical or logical organization of
the data files, although it is quite definitely true that certain formats are
much more amenable to the system than others. Files may be text or binary;
fixed record length or variable; binary files may have floating point values
stored in the native hardware format of the machine creating the data set.
Virtually any logical record structure can be accommodated, but
usually at the cost of support and accessibility to general users.
However, part of the mission of the PDS is to make this data available to the
community (now and in future), so there is a strong bias towards data
formats which can be easily read by many different systems. Consequently, the
great majority of PDS data sets will have these attributes:
- Fixed-length records
- One logical record per physical record
- Industry-standard storage formats (i.e., ASCII for text, IEEE for real
binary values)
Text Files
Text files, whether data tables, labels or documentation, have
only one hard and fast
requirement: the records must be delimited by both a carriage return and a line
feed. This ensures that the files can be read and displayed by every major
operating system at the cost of displaying some unrecognized control characters
on systems (like Unix) which do not require both.
While it is possible to write valid PDS labels for stream-format files
(i.e., text
files with variable-length records), this is most useful for labelling simple
documentation files rather than data files, where each field should be
defined and described in the label.
PDS recommends that data tables and PDS labels themselves be fixed-length
record files.
Binary Files
It is very difficult to deal with binary files without fixed-length records,
so variable-length binary records are very strongly discouraged. PDS does not
have hard requirements for binary storage format or byte order, preferring
instead to require documentation in the associated PDS label of these
attributes. Consequently, VAX users are not required to byte-swap their data
prior to submitting it to PDS, provided the data labels note that the bytes
are in the standard VAX (least significant byte first) order.
|
