Re: [Cfrg] should the CFRG really strive for consensus?

[Yoav Nir <[e-mail]>]

> On Dec 31, 2014, at 6:29 PM, Salz, Rich <[e-mail]> wrote:
> 
>> IRTF groups do not, technically, have to reach consensus. However, everyone does have to function on the same Internet at the end of the day.
> 
> Neither does the IETF.  As I recall, the phrase is rough consensus.  At least as regards X25519, I think it's pretty obvious that Microsoft is in the rough.

The consensus process is pretty terrible at choosing between two or more non-horrible alternatives. It’s fairly easy for relatively small groups to block consensus.

When it works, the consensus process produces the best solution that won’t get a group of participants to say, “This sucks, we’re taking out ball and going home.” In our terms that would be Microsoft declaring that the product of this process is not going into Explorer and IIS. 

Compare the process for the ChaCha20+Poly1305 draft. We submitted it, there were several suggestions for improving the text, one suggestion to improve the construction, which was accepted, and a couple of suggestions to modify the base algorithms that were not accepted and no big arguments came out of it. Contributions included better text, pseudo-code, links to research papers about the algorithms, and a more thorough set of test vectors. The group made the document better until there was consensus that it was done.

If the only proposal before CFRG was Curve25519, the process would be the same. I think the extra proposals caught the chairs by surprise. But regardless, it exposed how bad the consensus process is at choosing. For extra fun, see two similar cases at the IETF in the IPsecME working group. 

And now I should  add a sentence that starts with, “And the key to fixing this is…”, but I can’t. As long as we all want to interoperate on the same Internet, and as long as we don’t have any format authority (the way NIST does), we are forced to do our best to avoid people “taking their ball and going home”.  I don’t know how to fix this.

Yoav