[tahoe-dev] "Elk Point" design for mutable, add-only, and immutable files

[David-Sarah Hopwood]

Shawn Willden wrote:
> On Tuesday 15 September 2009 09:57:15 pm David-Sarah Hopwood wrote:
>> <http://jacaranda.org/tahoe/mutable-addonly-elkpoint-3.svg>
>>
>> I'll explain it in more detail tomorrow. I dropped the ability to have
>> write-only caps that do not allow reading, so only needed symmetric
>> encryption.
> 
> One thing that has confused me about your diagrams is the role of V in the 
> signature operation that generates "a SigC".  It says that V is the "key", 
> which I would think means it's the signing key -- but what, then is KC_sign 
> for?
> 
> What makes sense to me is that V = KC_verify and KC_sign are an asymmetric key 
> pair, and that the signing operation does not really involve V.  If that's 
> correct, I think you should move the 'key' label to the arrow from KC_sign 
> and remove the arrow from V.

Oops, you're right, that's a mistake in the diagram. Now fixed (at
the same URLs, since this was just an error rather than a new version
of the protocol).

> It might also be clearer to change the label on 
> V to KC_verify, and eliminate the "V = KC_verify" text.

It is called V in both the mutable and immutable protocols because it
plays the same role in each protocol. Otherwise it would be necessary
to use "KC_verify or UEBhash" when saying something that applies to
both protocols.

> Another relationship that perhaps should be made more clear is the 
> relationship between S and KR_sign.  If I understand it correctly, they are 
> another asymmetric key pair.

Yes. S = KR_verify; I didn't change its name because that would have
obscured the similarity to the previous version.

> However, your comment that you only need symmetric encryption for this version 
> would indicate that I don't understand any of what I think I understand, 
> because I don't see how you can accomplish the goals of either Sig_KR or a 
> SigC, as I understand them, without asymmetric encryption.

They use signature algorithms, which are asymmetric cryptography, but
not asymmetric encryption.

-- 
David-Sarah Hopwood  ⚥  http://davidsarah.livejournal.com