Strategy Handling in XIA2
-------------------------

Graeme Winter
9/AUG/07

The Problem
-----------

We want to be able to use xia2 - and keep with the principles of xia2 - to
help with data collection. This will have to include ideas like the elimination
of solutions, the constant idea that you may be wrong and robust and reliable
procedures which have the traps designed out.

The Idea
--------

To extend the principles of xia2 in the direction of data collection. This
can be done by for instance selection of three images rather than two for the
crystal characterisation, allowing for the idea that any of the autoindexing
lattices which come back may be right, and that any pointgroup for these
may be right. Also, we need to allow for the idea that at this time we don't
know but that we want to know as soon as possible.

Therefore I have this idea - we collect three images, autoindex, integrate
them and then use this as input for the strategy program. Each pointgroup
for each lattice should have a strategy calculated (may involve running
othercell and doing some reindexing of the matrix and the reflections...)
and then we see which bit crops up (e.g. 10 degrees) in all of the strategies.
We collect this bit *first* and use this to characterise more accurately the
crystal but *do not eliminate* possibilities here. Then we keep on going
with the data collection for the highest remaining possibility and so
on until we have data that the rest of xia is happy with.

Also need to bear in mind how this will relate to measuring data for
MAD / SAD experiments, how this will change once you have high readout
detectors (e.g. Pilatus) etc.