I did. I.e. I went to an ordinary public (not private) high school in Finland. It wasn’t even one of the better ones. This was, what, 30 years ago, so my Latin certainly is a bit rusty. I still think it’s a shame that in many parts of the world one has to be extra privileged financially to be able to do this. For me, studying a “dead” language like Latin at that age opened up a whole new way of thinking about language - a skill I’ve found very useful later in life.
Brilliant, thank you for that, it is superb to have such contacts. These will undoubtedly help validate or otherwise and lets then look forward to the next PARSEC killer. it will be a ride now in decentralised networking I feel. I cannot wait.
This is what many miss, it’s not about us owning/creating/inventing etc. or being smart asses, it’s all about anybody moving the industry forward and those that cannot see that are just in the way, that’s where my patience is probably not as great as it should be.
In any case, these sound like just the ticket for some peer review and the guys will relish any improvements etc. that we can make. Thanks again for doing this important task.
Yeah, other countries are much more advanced than the UK in that respect. Most kids here learn ‘the umbrella of my aunt is on the table’ in French for two years and that’s that. Mind you, once you can speak Finnish everything else must be a walk in the park
We’re getting off-topic here, but Finnish was actually my third language.
Also, whether a language feels difficult depends on where you’re coming from. If you’re English, then Finnish feels difficult. If you’re Estonian, it doesn’t.
Do you think there could be a chance of a posibility that some of the brilliant people you did send the paper to could help with the review. I have some thoughts about what a great impact it could have if some highly regarded people could be signed on the paper as reviewers.
No one is born technically minded, but most people can learn. Try this excellent article which partly explains PARSEC in a simple way:
Absolutely, David. And no need to thank me. My actions are quite self-interested as the Safe Network’s success is in my (all of ours really) best interest
That’s the aim re: having some credentialed folks from academia and industry review it.
Not asking for names (unless it is ok to disclose), but could you provide some more information on what institutions are going to review?
Can someone please explain Common Coin and Concrete Coin?
Common coin: a device that nodes can use to generate a “random” binary value (true or false).
Every time a node flips the coin, all other nodes are guaranteed to observe the same value for their coin flip. Also, the value must be unpredictable, which means that no malicious node can know the outcome before a honest node has observed it.
Concrete coin: a weaker (i.e: easier to design) version of the same concept where the value only must be common and unpredictable a significant proportion of the times (for instance: 2/3). The higher, the better, but doesn’t need to be all the time.
If you’ve only got a concrete coin, you can use the 3 steps “concrete coin protocol” to get the same benefits as a common coin would give you.
This is an adaptation of Silvio Micali’s idea that he demonstrates in the synchronous case in “Byzantine Agreement, Made Trivial”.
Using this and other techniques, we’ve reduced the problem of ABFT to that of finding an asynchronous concrete coin.
Right now, we’ve got an almost asynchronous concrete coin, as shown in our paper, but we’re still on the lookout for a fully asynchronous concrete coin to make the maths more glamorous. We’ve made progress in that direction last week. Hoping for more this week
Ok, I may have answered a bit more than the original question… Oh well
Great answer, thanks! One more thing out of interest … I understand the value of 2/3 for the min number of honest nodes drops out of the maths, as you’ve explained elsewhere. Does the fact there are three steps (Forced true, Forced false, Genuine concrete coin flip) in the concrete coin protocol (slide 69) relate directly to this ratio - ie there are two certain steps and one random one - or is that a coincidence? I’m presuming it does, and that the 1/3 of random coin flips are the only chance the Byzantine nodes get to influence the proceedings?
No. It is a coincidence. There are 3 steps because it’s the minimum viable:
- you must be able to decide true at some point
- you must be able to decide false at some point
- you must be able to use your genuine concrete coin at some point to give the randomness that’s necessary for liveness, but you can’t decide at that stage because you know the concrete coin won’t always be common
And as any elegant maths, the minimum viable number of steps is what’s used in practice (btw, it’s not bragging: all the credit for this idea goes to Silvio Micali as mentioned earlier)
Edit: I originally misread your question. This answer was actually the answer to the question I thought you asked May still be interesting, so leaving it here.
That’s a very interesting question, and actually, the 2/3 ratio is totally arbitrary Any ratio that is significantly non zero would be sufficient to maintain Liveness.
Agreement is never in question because no decision is ever taken based on the outcome of a concrete coin. When a decision is taken during one of the “forced flips”, you know agreement will be reached at that same step and maintained during any subsequent step. The concrete coin simply guarantees that if agreement can’t be reached during the forced flips (for instance because the adversary is very powerful and manages to follow an optimal strategy where he keeps the honest nodes confused), it will sometimes be reached during a “genuine flip” step (and then be maintained)
This is a basic a question but I don’t understand why a coin is needed.
Suppose node A has cast a meta vote for B (true if it strongly sees B and false otherwise). If consensus is not reached about B why not just stop everything simply because there are too many byzantine nodes? Why doing additional steps with a coin that may force A to gossip something that it doesn’t see would help emergence of truth?
An honest node forced to lie about what it really sees may help getting consensus, but it won’t be on real truth but on an “alternative fact” instead.
The concrete coin never makes a honest node lie.
A node’s observation (for instance a true or false meta vote) is only their original guess for the value of the meta vote. They may change their mind if they see that > 1/3 nodes disagree, which proves that at least one honest node disagrees.
This means that we can end up in a tie situation where all honest nodes are split between true and false.
The concrete coin is only used in such tied situations to “break the cycle” of a possibly powerful malicious adversary working very hard to maintain the tie forever so no decision is ever reached.
Pierre answered partially, but the deeper reason is the FLP impossibility: impossibility of deterministic, asynchronous consensus. Basically, if we have no way of telling whether a node failed or is just delayed, it turns out that with Byzantine Fault Tolerance there are always possible initial states that make any algorithm never terminate. This can be worked around by introducing indeterminism, as it essentially means that you are allowed to jump to an arbitrary state at some point, possibly breaking the infinite stall.
And we also prove that any agreement we reach will be on a value that was proposed by an honest node - so no matter what we decide eventually, someone honest has seen it that way.
So no-one found the Easter Egg? Ah well, I guess the prize wasn’t too enticing. Shame cos I think you’d like it (unless you happen to be supermax67). Here’s a subtle clue.
Yea I think when we all saw that in the video we thought “unusual must be some gimmick”. But being unable to understand the language and not needed for the understanding of the video I put it aside.
Definitely liked it @Nigel @Eureka18! We’ve always had a fairly open approach to presentations here - everyone has different views/preferences when it comes to their own talks/slides so we try to protect that element of personal choice as far as possible without becoming overbearingly corporate etc. So saying - totally in agreement that it never hurts to have SAFE Network branding on slides (other than full picture slides).
Unfortunately given all that’s going on at the moment, looks like we’re pressed for time re going back to re-record/rebuild this presentation with those slides this week (think Pierre had 80-odd whilst @Eureka18 has 30 or so as a basis for building out the other 50 or so). Another consideration re impact is that if we were to replace the video with a new upload, we’d lose those c.1000 views so far on YouTube - which would definitely be a pity. But again: we’ll definitely add it into the mix of tasks and see what we can do/whether it’s possible. The talk will be given again in different formats over the next few months so ideas can be integrated there as well so it’s valuable either way. So - thanks again for all the work @Eureka18!