Interesting, thanks for the link. Seems we arrived at similar conclusions regarding the hash function, with similar frustrations with respect to blake2b/2s.Funny that it's also for the same merkle set application. We ended up with a Merkle Mountain Range [1] instead of a Patricia tree. The MMR is append-only and makes pruning easy, which works well for MimbleWimble. You can navigate down the MMR with just the position the element was inserted at, so we just keep a simple index for that. Memory layout is great as a lot of it is immutable and sit close together, although the current impl doesn't leverage that too well yet. Wonder if Bram looked at MMRs? Patricia trees may make more sense for Bitcoin though.Proof of positions are cool, might look at that some more in the near future, when we're less busy implementing everything else ;-)-------- Original Message --------Subject: Re: [Mimblewimble] Switch to Blake2Local Time: July 24, 2017 6:44 PMUTC Time: July 24, 2017 6:44 PMFrom: kanzure@gmail.comTo: Ignotus Peverell <igno.peverell@protonmail.com>, Bram Cohen <bram@bittorrent.com>, Bryan Bishop <kanzure@gmail.com> On Fri, Jul 21, 2017 at 1:12 PM, Ignotus Peverell <igno.peverell@protonmail.com> wrote:So I'm considering a switch to the Blake2 [3] hash function.Bram recently made some comments about blake a few weeks ago: