Re: [Cfrg] I-D Action: draft-irtf-cfrg-argon2-10.txt

[Colin Perkins <[e-mail]>]

Hi,

> On 9 Apr 2020, at 12:22, Stephen Farrell <[e-mail]> wrote:
> On 09/04/2020 10:05, Milan Broz wrote:
>> On 07/04/2020 01:41, Stephen Farrell wrote:
>>> I did an IRSG review for -09. This addresses all the issues
>>> that I found there, except one, and that being (I think)
>>> the most important one;-)
>>> 
>>> The issue: What is a "primary variant" and what is an
>>> implementer supposed to do?
>>> 
>>> The above is a quote from my review of -09. Apologies if I
>>> missed some change in -10 that addresses this.
>>> 
>>> Let me make a suggestion: state that argon2id is mandatory
>>> to implement, and that the other variants are not.
>> 
>> Argon2id is combination of Argon2i + Argon2d, so implementing
>> all three variants should be quite easy (code must be there anyway,
>> it is just about providing external interface to it).
>> 
>> I think all variants should be mandatory...  and as I understand
>> the current RFC draft (3.1 section), it already says so:
>> 
>> o  Type y of Argon2: MUST be 0 for Argon2d, 1 for Argon2i, 2 for
>>      Argon2id.
>> 
>> (We use Argon2 in cryptsetup/LUKS2 and Argon2i (not id) is
>> currently the primary variant for the key derivation.
>> For compatibility with existing devices we need support
>> for all three variants. But obviously, this is just one use case,
>> RFC is more generic.)
> 
> I'd also be fine with all being mandatory to implement
> but would prefer just one. Even if there's not much cost,
> in terms of code, there would be follow on cost in terms
> of assigning code-points at higher layers, and in terms
> of the arguments that will ensue about which code-point
> to use/assign in various contexts. Providing more options
> at this level is usually a bad plan that's unfortunately
> attractive to many cryptographers;-)
> 
> But my main ask is just that it be clear what's expected
> of an implementer, so we get better interop.


I agree with Stephen that the draft needs to be more explicit about which variants need to be implemented. 

-- 
Colin Perkins
https://csperkins.org/