update to arkworks 0.4.0

[mimoo]

https://github.com/zksecurity/noname/pull/137 had to downgrade to 0.3.0, let's try to move to arkworks 0.4.0!

[Autoparallel]

I could take a stab at this!

[Autoparallel]

Curious what your thoughts are:

The naive approach of "just increment the version" yields some errors that point to the fact that kimchi may need a bump as well, unless we want to create some wrappers (which could work, but seems a bit iffy).

Namely: https://github.com/zksecurity/noname/blob/cf9af0b07c248561985fa13107d267655e3619de/src/backends/kimchi/mod.rs#L35-L42

We'll get:

the trait bound `ark_ff::fields::models::Fp256<kimchi::mina_curves::pasta::fields::FpParameters>: ark_ff::PrimeField` is not satisfied

when we try impl BackendField for VestaField {}

Stepping into kimchi, it's using ark_ff::PrimeField from 0.3.0 and implements PrimeField on Fp256 via the macro:

impl_Fp!(
    Fp256,
    Fp256Parameters,
    BigInteger256,
    BigInteger256,
    4,
    "256"
);

Is it worth trying to push up to 0.4.0 inside of kimchi prior to making the change here? Do we instead want to do:

pub struct VestaField(kimchi::mina_curves::pasta::Fp);

and implement the necessary traits on this wrapper? This solution seems less useful as it will likely just be undone if kimchi bumps arkworks.

[katat]

Nice analysis @Autoparallel !

I think we don't know when kimchi will bump the arkworks version. The wrapper approach you described seems to be way to go. If this wrapper works well, it would become the "template" solution to similar issues in the future as we may add new backend dependencies which we don't have control of.

How many traits does it have to implement? (We may need to evaluate if it is too much work to do so also.)

[Autoparallel]

@katat, I don't have the number of traits to impl off the top of my head, but I don't think it is too many. Also, they will likely be pretty easy to implement. My worry would be a bump in kimchi leads to undoing all of this.

If possible, poking at kimchi itself to push for a bump may alleviate more pain for everyone down the road. I'm happy to go there and chat about it too!

[mimoo]

there was a PR for that but it was never merged: https://github.com/o1-labs/proof-systems/pull/998

[Autoparallel]

Oof, and it is quite old.

Maybe it's safe to say that there is little initiative to get that bumped upstream then?

[mimoo]

yeah. I'm not sure what's best to do here... We might want to either:

• give up on this for now
• or push folks on the kimchi side to update their version of arkworks (cc @bkase)
• or start deprecating kimchi

[Autoparallel]

@mimoo what would be the replacement for kimchi?

[mimoo]

There's no real replacement for kimchi, kimchi is one backend among other backends. I think we should punt on this decision and pause this issue at the moment