RE: [boost] Loki SmartPtr study: Policy orthogonality issues

From: Aleksey Gurtovoy (alexy_at_[hidden])
Date: 2002-01-07 11:04:59

David Abrahams wrote:
> I think you've lost the context of my remark. I was trying to 
> explore the question, "given that we all agree [;-)] that a DSL 
> is too limiting for our purposes, how can we make a 
> policy-based design that is easiest to use?"

I would say that forcing the users to (always) specify the component
characteristics by means of ICCL is too limiting. For one, you have to
always specify all the implementation parts of a component being generated,
including the ones you don't care/have no knowledge to make an intelligent
decision about. Also, it's often a case that some important for the user
characteristics of the component just can't be directly expressed using ICCL
vocabulary. For example, if you are exposed to the policy-based
do-it-yourself smart pointer configuration interface only, you cannot choose
a specific smart pointer implementation basing (among other criteria) on the
pointer's size, and I consider this to be a severe limitation of the
approach. 

FWIW, I think that section 5.9.7 ("Configuration DSLs") of the book does a
good job of explaining the purpose and importance of DSLs.

Aleksey