Re: [keyassure] Opening issue #21: "Need to specify which crypto

[Paul Hoffman <[e-mail]>]

On 3/4/11 4:13 AM, Andrew Sullivan wrote:
> On Fri, Mar 04, 2011 at 11:51:10PM +1300, Peter Gutmann wrote:
>> is finalised all I need to do is plug the algorithm in.  Putting the same
>> thing into DANE isn't hard, and makes it easier to deploy future-compatible
>> implementations from day one.
>
> I don't see any reason to say specifically, "We'll be moving to _x_ so
> you better have a placeholder for it".  That will simply encourage
> foolish programmers to write one placeholder spot, and then when SHA-4
> or whatever comes along we have a new problem.
>
> The right way to do this is something like the unknown RRTYPE in DNS.
> In that case, you have to be able to handle anything you get (within
> the valid codepoint range) _even if_ you don't understand it.  By
> "handle", I mean, "be able to pass messages; not spit up and die," and
> so on.  It does not mean, of course, that you have to be able to do
> whatever additional section processing or whatever is required, and
> you don't need to be able to do anything useful (if you don't know
> what NSEC3 is, you're not going to be able to validate with it).
>
> I'm not sure how to say something similar in this case, but I think
> it's the right sort of framework.

The way I read Warren's proposal, we get exactly what you want. There 
are two different mandatory code points defined for two different 
hashes, so we are sure that the deployed code will be able to handle 
hash agility. We don't care which algorithm people use in practice 
because they are both mandatory, so there is no spitting up or dying.

Does that work for you?

--Paul Hoffman