Introduction

Asterix applications currently perform their data interface using a collection of routines in ASTLIB (BDA,LIST,SSO,PSF,HIST to name the major ones) and low-level access calls in the native file representation (HDS). The number of instances of the latter is decreasing asymptotically as extra abstract routines are introduced. The major area in which low-level access is still performed is in the the access of pointing information and other header components.

Shortcomings in the interface routines are now becoming apparent on a number of fronts.

• FITS. Support for at least a well defined subset of the plethora of FITS formats will be essential if Asterix is to survive in the medium term. FITS is the only sensible machine-independent format for storing event data.

• NDF support. We gradually extend the BDA system to cope with the odd dataset from KAPPA etc. which uses aspects of the NDF definition we'd previously ignored. This has been done in a rather ad hoc fashion with the result (for example) that DATA_ARRAY.DATA is supported, but not VARIANCE.DATA.

• Simple programming interface. The commonly used interface routines in Asterix number between 100 and 150 which is far too many for a part-time programmer to be aquainted with, and even leads to professionals rewriting code because they are not aware that a particular function exists. Even if we could gather together all our interface routines and write a big document, we'd hardly dare call it a homogenous set with a common design philosophy running throughout.