Transforming the Digital Object

A digital object can be transformed in many ways. The question is which transformation is acceptable in some ways. OAIS uses the term Transformational Information Properties which should be maintained over any transformation. For example, when transforming scientific data one should not lose significant digits in numbers, or bit patterns with special meanings e.g. quality flags, should be dealt with carefully. The semantic relationships between the parts of the original file must also be maintained. See Transformational Information Propertiesfor further details.

When the transformation is carried out the values of the Transformation Information Properties should be checked and then details recorded in the Provenance Information of the new Data Object.

If there are no explicit Transformational Information Properties supplied then scientific judgement should be used. As mentioned above one would require that the Transformation must maintain the accuracy and precision of the object being Transformed. For example one would not transform a file containing IEEE 64 bit reals into 16 bit signed integers because the numerical range of 16 bit integers (-32768 to +32767) is much smaller than that of IEEE 64 bit reals (+/-10**(308) ) and numerical precision would be lost.

A complex object such as a FITS file with multiple images and tables could be transformed into a directory containing separate images and tables. Alternatively it could be Transformed into an HDF5 file. Special care would have to be taken with the FITS file headers, and the relationship between components.

To Transform a FITS file to HDF5 the application fits2hdf (https://fits2hdf.readthedocs.io/en/latest/) may be adequate; alternatively there are a number of software libraries, for example https://fits.gsfc.nasa.gov/fits_libraries.html.

Transforming a table to a CSV file is possible but special care must be taken with units, semantics and formulae.

Suitable examples of such Transformations, as well as Functions created previously, may be available.

Tools for Transforming a Data Object

A Data Object may be Transformed using applications such as fits2hdf using the command line or a script. The application may provide its own GUI. The application may also be integrated into the Functionsin LABDRIVE.

Changes to the AIP

Representation Information

The appropriate Representation Information must be associated with the new Data Object noting that the Structural Representation Information and Other Representation Information will almost certainly be completely changed whereas the Semantic Representation may have parts, such as data dictionaries or ontologies, which are applicable to the new object.

Provenance

Details of the application may be inserted automatically in the headers of the new Data Object, in which case the Provenance Information should note this and provide details of how to obtain this information.

The checks of the Transformational Information Properties should be recorded in the Provenance Information.

Fixity

The hash code will certainly change.

Context

The Context Information should be unchanged.

Access Rights

Access Rights should be unchanged.

Other elements

The TIP and Presentation Objectives should be unchanged.