J. Wuttke suggested to define a hierarchy and a dictionary for the data which is compatible with various representations, e.g. wrapped YAML for ASCII representation and HDF5 for Binary representation.

This will allow to easily transform one representation into an other, and it avoids different dictionaries for different representations. I.e. the work is done only once and confusion is reduced.

Though this sounds reasonable there might be some problems with this:

• A feature of HDF5 is that an entry might have an attribute, where there is no equivalent in YAML.
• It might result in incompatibility with canSAS.

• The clientele for the ASCII and the binary representation are not identical and the different requirements result in different content of both.

  E.g. the ASCII representation contains much less data to stay human readable so that a transformation to the binary representation might not be useful anyway.

tasks

• Check the compatibility with canSAS
• Discussion on the necessity of transforming in both ways
• Collect arguments for or against the common structure (what does it mean for programmers? what about possible future formats?)

decision

• Do we insist on an attribute-free representation of the data?