From: Maarten Kronenburg (M.Kronenburg_at_[hidden])
Date: 2006-11-30 07:09:26
"Paul A Bristow" wrote in message
> >Note that besides Daryle's new effort, which I haven't looked
> >at, there's
> >Boost BigInt by Ron Garcia and friends in the sandbox as well.
> > Not sure why we keep reinventing this but never finishing...
> As, I suspect, a representative of those who don't fully understand the
details of these discussions, I feel impelled to say how
> very strange I find it that Boost has several flavours of fancy points but
lacks any sort of bigger integer.
> Surely, bigger integers are quite fundamental; and have many potential
> I ask again if our review process is partly to blame for this.
> IMO people are not going to put in the boring work on finishing it to the
rightly rigorous review standard unless they feel they
> have a good chance of getting it through (and maybe not even then - but
would be happy for someone else to do the drudge work on
> testing and documentation). Do we need some process for deciding that a
particular design/prototype is a 'candidate for work towards
> a full review' in order to provide that encouragement?
In my opinion software design is not mathematical proof that one design is
better than another, but there is the mathematical fact that the set of
unsigned integers is a subset of the integers, and the set of modular
integers is a subset of the integers. In the book "C++ Coding Standards" by
Herb Sutter and Andrei Alexandrescu, in chapter 64: "Blending static and
dynamic polymorphism judiciously", it reads:
"Due to its characteristics, dynamic polymorphism in C++ is best at: -
Uniform manipulation based on superset/subset relationships: Different
classes that hold a superset/subset (base/derived) relationship can be
In my "design decisions" I will quote this passage literally. Of course
anyone is free to propose a different design, and then the LWG would have to
make a decision which design to prefer.