[Cryptech Tech] SHA-3 and Ed25519

[Павел Шатов]

On 05.03.2015 16:35, Joachim Strömbergson wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
>
> Aloha!
>
> Павел Шатов wrote:
>> http://opencores.org/project,sha3
>>
>> Can we use this?
>
> Possibly. The license is Apache. The cores are fairly big compared to
> other SHA-3 implementations.

What do you mean when you say big? This SHA-3 core is 500 lines of code, 
your sha512_core.v is about the same (600). Synthesis results are again 
somewhat similar to your SHA-512 core (4500 LUTs for SHA-3, 4400 LUTs 
for your SHA-512).

> But if Bernd has a core too I would love to know more about it. Bernd is
> part of Cryptech and (at least I) trust him to do good stuff. That is
> why I asked him specifically.

I agree, if we have our own core, it's better to use that, of course.

> We are aware of the OpenCores project. I have spent a fair amount of
> time with the project in its early years and I have tried quite a few of
> the cores. There are some really good coresin OpenCores, OpenRISC with
> support cores for example. But there is also a huge number of broken
> cores, bad cores, half finished cores etc in there. Most times when I've
> looked at OpenCores I've ended up disappointed.

Yes, there are way too many junk projects at OpenCores. SHA-3 should be 
OK though. It is certified and its author has other crypto cores 
published there (AES and Elliptic Curves), he should be pretty competent 
in what he is doing.

> There is also a general discussion we should have regarding "core"-cores
> and what we should maintain. We can build something fairly fast by
> bringing in cores from external sources. But if we don't maintain them,
> can we and Cryptech users trust them.
>
> In the case of the SHA-3 core we might be able to use the core on
> OpenCores right now. But if/when SHA-3 becomes a vital function in our
> use cases we might (I think) have the source in our repo.
>
> If we for example decided to replace SHA-512 in the RNG with SHA-3 I
> would definitely have SHA-3 as part of our code base.
>
> Makes sense?

I don't quite understand this. Does Apache license allow us to have a 
copy of SHA-3 source code in our repository retaining all the original 
copyright information or not?

What maintenance are you taking about? According to information on 
OpenCores their SHA-3 was updated more then 2 years ago, the source code 
should be stable.

--
With best regards,
Pavel Shatov