[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: GNUnet PoWs hashfunction
From: |
Schanzenbach, Martin |
Subject: |
Re: GNUnet PoWs hashfunction |
Date: |
Sat, 18 Apr 2020 13:35:58 +0200 |
> On 18. Apr 2020, at 13:24, Jeff Burdges <address@hidden> wrote:
>
>
>
>> On 18 Apr 2020, at 09:28, Schanzenbach, Martin <address@hidden> wrote:
>> Ideally, the PoW should take ~1 week. I am not sure yet if the chosen
>> parameters fit this requirements.
>> If you have any input on parameter choice or can run a revocation (with the
>> argon2 code from that branch) on your device with 17-25 WORKBITS that would
>> help greatly.
>> You will need to have libargon2 installed.
>
> Is 1 week at 100% CPU right? Any parameters sound painful at such durations.
> Is this for revoking once you lost the key or something? I though PoW only
> exists here because you broadcast? Or is that 1 week as a very backgrounded
> task?
Yes, I think we actually only have the PoW in order to prevent mass flooding of
revoked keys.
1 week 100% for a single core CPU was the past implementation.
>
> As an aside, there is a verifiable delay function (VDF) project by the
> ethereum foundation, which protect against parallelism. I’d expect VDFs
> provide less protection against ASICs per se than Argon2, but they should
> ideally provide relatively high confidence in ASIC speed. VDFs should only
> consume one core while running, which might improve user experience. We can
> chat about this once some VDF achieves the desired confidence in ASIC speed.
> ;)
>
> Jeff
>
signature.asc
Description: Message signed with OpenPGP