In article <87y979kq32.fsf@snark.piermont.com>,
Perry E. Metzger <perry@piermont.com> wrote:
> 
> No. Thor is absolutely correct. There is no advantage to an MD5
> checksum of the file over any random function yielding 128 bits if you
> do not check the hash later on.

So long as the "function yielding 128 bits" is truly a *function of the
data*.  A random number generator destroys the important property that
identical source run through identical toolchain bits (which should
include prebinding) produce identical files.  Violating this property is
a major PITA, especially for testing...  IIRC, GCC used to have to have
(maybe still does?) cruft to deal with comparing stage2 and stage3 on
platforms where the toolchain inserted random values (timestamps, for
example) in object files.  There's no reason to put anyone else through
that kind of nonsense. 

Wether the function is MD5 or CRC128 or whatever, I don't care.  Given
that we already *have* code for MD5, it seems like a reasonable choice. 
A type field is also an excellent idea.