Comments on: What is Practical Byzantine Fault Tolerance? Complete Beginner’s Guide

By: Amit Vishwakarma

Amit Vishwakarma — Mon, 18 Nov 2019 16:06:35 +0000

How PBFT is achieved in Hyperledger fabric ?

By: Joe Blow

Joe Blow — Wed, 28 Aug 2019 10:31:33 +0000

In reply to ldc. Yes, I noticed the same thing. I am not a smart guy, so I read the thing ten times. :D It should say, where (n - 1/3) of the nodes are NOT malicious.

By: Mental Nomad

Mental Nomad — Fri, 01 Feb 2019 16:55:09 +0000

It’s not about a 35th reply… you’re throwing yourself off by thinking about the next reply. Back up for a moment:

If f = 33, then AT MOST 33 people in the whole group of 100 are liars.

When you receive 34 replies, then you have AT LEAST one honest person in your group of replies, even if (by chance) the other 33 are all liars.

IF your 34 replies are ALL the same, and AT LEAST one of them is honest, then all of those 34 replies must be giving the honest reply: they all match each other, and you know one of them is honest. You now know, for sure, what the honest reply is.

It makes no difference what a “next” reply would say – you already know what the honest answer must be. It isn’t impossible that a 35th answer comes in, and it may be lying… but if it is lying, you’d know, because you already know the honest answer from having already received 34 matching answers!

Of course, all of the above is only true so long as there aren’t actually more dishonest actors than f… and if your initial group of 34 replies are NOT all the same, then you know some of them are lying, but at least you’re aware that someone is lying and that you don’t yet know the truth.

By: ByzantineMan

ByzantineMan — Fri, 01 Feb 2019 01:35:13 +0000

” If you get f+1 replies which are all the same, and you know that at least one was honest, then you know they’re all honest.”

How is that true? If there are 100 people in a group, and f is 33, then if I get 33 + 1 = 34 replies that are all the same, why is it impossible that the 35th reply couldn’t be a liar?

It seems logical that I would have to receive 2f+1 replies before I can know they are ALL honest, because that would be the inversion of f.

Please explain.

Great article by the way!

By: Mental Nomad

Mental Nomad — Fri, 30 Nov 2018 19:16:45 +0000

No, they only need to wait for f + 1.

If there are maximum f liars, and you hear back with more than f replies, at least one reply will be honest. If you get f+1 replies which are all the same, and you know that at least one was honest, then you know they’re all honest.

(3f+1) describes the size of the total population in terms of f, the number of faulty nodes. The honest population must be more than double the faulty population for this to work… so if there are f faulty, the honest must number (2f + 1), and the total population is (3f + 1).

By: Mental Nomad

Mental Nomad — Fri, 30 Nov 2018 19:09:27 +0000

In reply to shah. Shah has it right - should be (n-1)/3... the text may have been improperly formatted after copying from this kind of format: n-1 ___ 3

By: shah

shah — Wed, 24 Oct 2018 19:33:46 +0000

In reply to ldc. (n-1)/3

By: John

John — Sun, 05 Aug 2018 05:22:16 +0000

>> client awaits f + 1 (f represents the maximum number of nodes that may be faulty) replies from different nodes with the same result.

Is it 3f+1 (to satisfy the 2/3 rd rule)?

By: ldc

ldc — Tue, 10 Jul 2018 08:17:21 +0000

… as long as at most `(n-⅓)` are malicious or faulty at the same time.

n minus ⅓ ?????