[Quantlib-dev] Shortish-term plans?

[Luigi B. <lui...@fa...>]

Hi all,
	while I'm happy with the recent trend of quarterly releases  
(with the addition of the occasional bug-fix release :) I've been  
thinking whether we should aim at increasing the minor version number  
as well.

As the Monte Carlo framework is the part that most looks like being  
near a decent status, I think it could be a good candidate for making  
its interface stable, wrapping it up and calling it 0.4.0.

However, the MC framework does have a number of open issues to be  
solved before calling it stable, such as:

- the fact that the Path class stores drift and diffusion instead of  
the asset value. Apart from the inconvenience of having to reconstruct  
prices, this has the shortcoming that it is currently hard-coded in the  
path generator (and the path) that the given stochastic process  
describes the evolution of log(s). Therefore, processes such as square  
root cannot be used.

- some sort of multi-diffusion process should be defined so that the  
single diffusion processes and their correlation could be stored  
together in one structure.

- I'm sure Nando will have quite a few suggestion for further  
additions. The ones that imply an interface change should be singled  
out from the ones that can simply be implemented later on top of the  
basic framework, so that we don't set outselves a nice but impossible  
goal for the milestone :) This might also apply to Neil's least-square  
Monte Carlo, whose stabilization could be postponed.

Toughts?

Later,
	Luigi