AXIS values correspond to bin centres, and will generally be stored either as spaced arrays (for regularly spaced values) or primitive arrays. REGULARLY SPACED VALUES SHOULD BE STORED AS SPACED ARRAYS SO THAT SOFTWARE WILL BE ABLE TO RECOGNISE THEM AS SUCH. Software finding a primitive array of axis values should assume that it could be irregular and act accordingly (if it can't handle irregular spacing it must warn the user and either proceed assuming regular spacing, or exit).
NORMALISED=.TRUE. indicates that the values in the DATA_ARRAY have been normalised with respect to the axis values of the axis in question (e.g. count/s or count/arcmin**2). Note that if NORMALISED is not present it is assumed that the data are NOT normalised.
WIDTH is the bin width, which may be scalar (all bins of equal width) or vector (one value for each value in the axis array). The bins are assumed to be centred on the axis values. If we find a requirement for offset bins, or for bin durations which are not equal to bin widths, then we will have to make use of the proposed AXIS extensions, to represent the extra information.