I have been toying with the idea of a public domain set of standards
for a set of classes for the last two years, and I think several things
need to be settled by means of public discussion first. I would
propose that a standard set of objects needs to be extensible not just
in terms of additional objects, but in the real ability to extend the
functionality Along with the ability to curb those extensions so that
an individual task does not wildly overwhelm the systems capability.
For example, good objects for such a system might be able to be used
for a program simply looking at the structure, in which case knowing
ribbon structures and secondary structure information might be quite
valuable, whereas individual charge distributions for the residues
might not be so valuable, but would be critical for any energy
calculations one might want from another application. Another point I
would forward for discussion is that it might make sense to build the
objects from a set of base classes that are not specific to a single
implementation of a container class, but rather at a lower level such
that the objects are highly portable between C++ implementations. Ah,
if I had 6 months free I would swim in this project, but grad school
goes on and it slowly percolates in some small backwater of my mind
:^)
Tom Branham
branham at binah.cc.brandeis.edu
