[Dovecot] [OT: bzero()] Re: password generation tool and additional hashes

[David Champion]

* On 2004.07.27, in <4105FE76.1000508 at lxnt.info>,
*	"Alexander Sabourenkov" <screwdriver at lxnt.info> wrote:
> 
> HISTORY
>      A bzero() function appeared in 4.3BSD.  Its prototype existed 
>      previously
>      in <string.h> before it was moved to <strings.h> for IEEE Std 
>      1003.1-2001
>      (``POSIX.1'') compliance.

bzero(), bcopy(), and bcmp() date from early days of C/Unix
standardization, when everyone had one. These were the BSD byte range
operations, and memset(), memcpy(), and memcmp() were the non-BSD ones.
bzero() maps precisely to memset(), and bcmp() to memcmp(), provided you
swap appropriate arguments. These days, since mem*() are in the standard
library and b*() are not, it's generally wiser/more portable to use
them. bcopy() is slightly different from memcpy(), but if you have no
need of this difference, it's likewise better to use memcpy(). b*() are
not necessarily supported on a given platform; mem*() are.

The difference between bcopy() and memcpy() is narrow, and occurs only
when the destination address plus the range is greater than the source
address -- i.e., the copy operation will overlap its own source. In this
case, bcopy() is guaranteed to use at least two copy operations, thus
preserving the original data. Memcpy() can use only one copy operation,
overwriting the original data. (Or: bcopy() is focussed on accuracy, and
memcpy() may be optimized for execution speed.) This distinction tends
to bring the other b*() and mem*() operations with it, as they're all
parts of families.

That's why you'll find it deprecated by some, and praised by others.

-- 
 -D.    dgc at uchicago.edu                                  NSIT::ENSS