multiple file signing oddness

[David Shaw]

On Sun, May 23, 2004 at 05:53:29PM +0100, Nicholas Cole wrote:
>  --- David Shaw <dshaw at jabberwocky.com> wrote: 
> > Very interesting.  That's not ideal since the syntax
> > that is used to
> > generate the sigfile should be usable to verify the
> > sigfile.
> > 
> > Changing it in 1.2.x is not a good idea as it may
> > break some
> > assumptions and/or scripts, but I'll fix it for
> > 1.3.x.
> 
> Dear David,
> 
> Thanks for looking in to that.  I am right that there
> is nothing in the signature itself to tell a recipient
> in which order the files were signed, am I?  In that
> case should gpg 'search' for an order in which the
> files verify, or would that add too much overhead?

It should definitely not search.  When you sign multiple files that
way, you are effectively signing the concatenation of the files.

So if a.txt is "foo" and b.txt is "bar", and c.txt is "foobar":

   gpg --detach-sign a.txt b.txt

is identical to:

   gpg --detach-sign c.txt

If GnuPG searched for an answer, then checking the signature against
b.txt & a.txt ("barfoo") would be valid, and it should not be.

David