Re: OT: Number bases

On Wed, Feb 11, 2009 at 7:03 AM, Henrik Theiling <theiling@...> wrote:
> Ternary logic would also eliminate those awful hacks from computer
> languages: unsigned numbers.  They are a major cause for serious bugs
> like buffer overflows and integer underflows.
Natural and nonnegative numbers weren't invented for the convenience
of programmers, ya know.  There's a bit of precomputer history there.
:)

Of course the primary motivation for unsigned integer types in
computer programming is space savings.  By shifting the range from
(roughly) centered on 0 to having a minimum at 0, you double the
maximum value without using any more space, at the cost of not having
negative numbers available - which for many applications is not a
handicap at all.  My weight in pounds fits in an unsigned byte; to fit
it into a signed byte I have to convert to kg. :)

> Humans just don't care whether a number is signed or unsigned (and rightly so), so
> programmers make mistakes.
But humans also sometimes don't care whether a number is positive or
negative.   Two things here...

1) Conflating programmer interface with hardware architecture.  Any
high-level language should have support available for
arbitrarily-large-magnitude integers, rationals, arbitrarily-precise
approximations to reals, complex numbers, etc.  But those don't all
need to correspond to native machine types.

2) Solving the wrong problem.  As a programmer, sometimes I want
sqrt(-1) to quietly return i so I can go about my  business; at other
times I want it to scream bloody murder because a complex answer
doesn't make any sense.  Knowing the difference is hard; making the
language and/or hardware able to deal with complex numbers doesn't
solve the programming problem.  At worst, it hides the programming
problem so that it goes unnoticed until a bug is tripped.

--
Mark J. Reed <markjreed@...>

Reply