Start define requirements for a/the FADS

Preben Grosbol (
Mon, 08 Sep 1997 13:41:21 +0200

Dear FADS participants,

From the last years discussion it seems clear that many image
processing users would like to get away from using the current large
monolithic packages and move to a more modular system.

A main reason being that one frequently needs several packages to
perform an optimal reduction and analysis. This forces the user to
learn several different command languages without any real gain as
they all do almost the same, just differently.

What the average user would like (assuming that I'm such a one) is to
have his prefered environment (any of several) which can execute any
of a large set of applications. For this to work one would need
several things to happen:

1) A true separation between the user environment (client) and
applications (server), and
2) a common client/server interface definition for astronomy.
This must include both the technical interface (such as
the Beans in Java) and an astronomical definition including
conventions on units etc.

Actually, one would argue for a de-regulation of astronomical
software i.e. the development of user environments and astronomical
applications should be strictly separated.

Thus, one should be serious about it and start to set up the
requirements for such an interface from a true astronomical view
point, and not just meet once a year and say something should be done.

Of course there is a reason for pushing this issue. During the
analysis of ESO VLT Data Flow System, it became clear that a true
client/server relationship would be essential for the VLT Pipeline.
Although we will use a specific Data Reduction System for our first
implementation of the VLT Pipeline, it is also clear that no current
system can be expected to survive through the operational life of the
VLT (i.e. 25-50 years). The reduction packages for the VLT
instruments will be made by different instrument consortia but we want
to reuse as much as possible from one instrument to the next. The
reduction packages should work in the ESO pipeline but also be usable
by external users who want to re-reduce their data.

These and other arguments have lead us (ESO/DMD/DPG) to start an
internal discussion on how to define reduction entities (we call them
Reduction Blocks) for the VLT Pipeline, independent of any special
reduction system. We would welcome a more broad discussion of this
topic and hope that this years FADS BoF could start this process.

Best regards,

Preben Grosbol

Data Pipeline Group,
Data Management Division,
European Southern Observatory.