Small Vectors Proposal

[Walter Bright]

Andrei Alexandrescu (See Website For Email) wrote:
> It is a change that I personally think should be made. Walter devised 
> the one cast per class rule out of caution when facing an issue known to 
> be potentially disastrous. Generally, the "when in doubt, don't put it 
> in" approach turned out to be a winner for D in general --- the "length" 
> Chernobyl disaster notwithstanding :o).

Yes, it's the "when we don't know what the semantics of this combination 
of features should be, make it an error" rule. That way, when it 
sometimes eventually becomes clear what those semantics should be, we 
don't have a backwards compatibility problem.