Changing preferences

[David Shaw]

On Sep 21, 2008, at 11:57 PM, Robert J. Hansen wrote:

> David Shaw wrote:
>> If someone wants to know how to set their preference list, they're  
>> not
>> trying for new and fun ways to violate the spec.
>
> No, but they may be operating on the assumption their preference list
> matters.  (Which it very often doesn't; encrypting-to-self and another
> recipient means there's a 50/50 chance their preference list will be
> treated as a cap set.  It would appear this ought to be made clear in
> the docs.)

I'd welcome docs that make it clear, but I question how easily it  
could be made "clear" in something pithy enough for a man page.  If  
you make it simple enough to fit in the man page, you will get scolded  
for not covering some obscure case with v2 keys or something.  If you  
make it complete, it's too big for an already large man page.

I'd be content with something that says "List algorithms in the order  
in which you'd like to see them used.  If you don't include 3DES, GPG  
will add it automatically at the end.  Note that there are many  
factors that go into choosing an algorithm, and so GPG may or may not  
follow your chosen order for a given message.  However, it will only  
ever choose an algorithm that is on the list of every recipient key.   
See also the INTEROPERABILITY section."

> GnuPG's preference lists are arcane and counterintuitive, and the  
> source
> of a great deal of frustration.

If they are so horrible, suggest a different way to handle them.   
Better to fix it in code rather than document something you feel is  
confusing.

David