From: Peter Dimov (pdimov_at_[hidden])
Date: 2000-12-04 13:03:33
> > I may want to use std::vector<float, pool_alloc> in a library;
overriding
> > the global new[] is a per-project decision, not per-component.
>
> OK, I'm convinced. I'll allow sized_pool to use a user-defined allocator.
> The Standard allocator requirements are not suitable for this case; I'll
> have to invent new allocator requirements (hopefully under a different
name;
> "allocator" is way too overloaded here already). The semantic
requirements
> are all fairly obvious, except one: should the user-defined allocator be
> able to have state (instance-based)? I think that static functions should
> suffice. What do you think?
I agree.
IIRC it's an open issue whether the standard allocators are permitted to
have state - their instances must compare equal.
-- Peter Dimov Multi Media Ltd.