Re: GSL 2.0 roadmap

[Gerard Jungman <jungman@lanl.gov>]

Well, the discussion has started, and people are starting to step
out with their "2.0" ideas. The way I look at it, there are three
things that need to happen, in general terms:

  1) The backlog of 1.x-incompatible changes needs to be cleared.
     This corresponds to stuff on Brian's 2.0 todo list; probably
     lots of other stuff too. We need to make sure we overlook
     nothing.

  2) New modules can be considered for inclusion. 

  3) Changes to existing modules must be considered, where there are
     clear needs for improvements, in implementations, interfaces,
     or both.


Set (1) is fairly well-controlled, although it will take some
care to make sure everything is done right. I'm sure Brian is on
top of that. Set (2) is easy in a sense, since by definition
it involves well-encapsulated changes/additions.

Clearly, set (3) is the most difficult. Major issues here
are related to complex numbers, vector/matrix abstractions,
linear algebra, special functions, and FFTs. There are others,
but it will take some time to create a full assessment of the
existing modules. Detailed input from users will be needed.

However, we do need to decide on some plan for set (3), since
small changes in set (1) might affect or be affected by ideas
from (3). Some partial assessment of where we are is needed.

A GSL state-of-the-union is a large task. I have been thinking
about it, on and off, for months. In a followup message, I will
include my (incomplete) thoughts on this. These are almost all
critical thoughts, since I don't see any point in patting
ourselves on the back.

Feel free to blast away at it.

--
Gerard Jungman