Aaron,
My point is that you generate a different binary for each user --
the same code, but subtly different.
There are many ways to do this by perturbations of various
sorts.
Then you sign the binary with PGP.
Each binary is then different and signed,
so that you can verify that it is yours.
But the user has no way to determine whether it is
authentic.
Then they must get it from your site.
If they get it from somewhere else, it may be
a virus-infected copy.
The user's only authentication is the fact that
they get it from your authorized site.
Now if someone copies one of your authentic binaries and
distributes it, users do not know if it is one of the
virus-fitted binaries that you distribute into "incoming"
directories, or a good copy,a nd they would have to
completely disassemble the code and understand it in order
to know if it has a nasty bit of data-destroying code.
I.e., this methodology uses a kind of game theory.
The user takes a risk in buying pirate software.
When it becomes generally known that many Palm software
writers are usuing this methodology, users will
be terrified into getting their software from the source,
namely you, and they will gladly pay $10 for the certificate
or whatever.
Therefore your point about registration code generators is
not relevant. Just perturb a few function offsets in
your binary, so that the jump-vectors are perturbed.
Then you can generate a million different authentic binaries,
and a million bug-infested versions.
But because you ahve the secret PGP key, only you can
tell the difference.
Does this meet your objection?
Cheers,
Alan Kennington.
