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

From: Andrei Alexandrescu (andrewalex_at_[hidden])
Date: 2002-01-13 05:17:57

> 1)  What good is policy based programming if some of the best programmers
> in the world discuss for days what is best for a rudimentary class like
smart
> pointer.

This is a false assumption. Designing smart pointers is a very challenging
feat. The chapter in mc++d on smart pointers is the longest of all.

> Isn't it the point of /generic policy based programming to
> eliminate long drawn out decision making? I mean if the experts can't get
it
> right quickly, what chance does your typical programmer have? I think he
> would starve.

I think Loki does a reasonable job at making it easy for system designers to
implement and maintain their designs. Still many decisions belong to them.

> 2) What exactly is the defininition of a policy?  I mean precisely, what
are
> the criteria? There seems  to be this vague notion of it, but I don't see
a
> precise definition i.e. in EBN format or whatever. To me, precision is the
> difference between science and metascience.... understanding and
confusion.
> So if there is a definition I would appreciate someone pointing me to it.

EBN form? I think that wouldn't work. Informally, policies are simply
compile-time interfaces. There's no rigurous definition that I know of.

Andrei

---------------------------------
Check out THE C++ Seminar:  3 Days with 5 Experts
http://www.gotw.ca/cpp_seminar/