Quote from: Ix on June 02, 2018, 12:20:04 PM

But I trust my grandmother more than I trust CNN, or Coinbase, or the government.

This discussion really distills down to this point.
If 50+% of the stake has someone they trust with a live version of events and all of them have the same version of events,
then they can all choose the correct fork and the remaining minority of the stake can see the majority has decided on a fork.
Then the attacker loses.

But you would still have an additional problem in that to detect double-spending with 100% objectivity requires 100% finality of epochs.
IOW, all of you in the 50+% stake will not be able to agree with 100% certainty on a double-spend without 100% finality.
Yet 100% finality of epochs (as opposed to probabilistic finality of transaction confirmation) requires permissioned set of validators of which only 1/3 of them can stall the entire chain
and the only way to unstuck the (transaction confirmation of the) chain is to hardfork.
The attacker then only needs control of 1/3 of the validators in order to short the token and profit.
The chain's protocol can't confiscate the security deposits of the non-responding validators because they may be legitimately under DDoS attack or suffering from some general failure such as Amazon or Azure outage.

Also permissioned validators is a political clusterfsck as described in the EOS section of @anonymint's most recent blog:

https://steemit.com/cryptocurrency/@anonymint/scaling-decentralization-security-of-distributed-ledgers

A consensus system with nothing-at-stake permissionless validators can fake a network outage by Sybil attacking the validator set.
Your Decrits design apparently forces new validators to queue up and be approved by many epochs before joining or leaving, but this is in essence a permissioned system,
because then 1/3 of the validators can stop the forward movement of the chain and those queued validators never become approved.
If you use an elapsed time instead of epochs, then that opens a different sort of security hole.

Quote from: Ix on June 02, 2018, 12:20:04 PM

Regardless of whether or not objectivity exists, there will be a loss to those attacking the network (and to those actively defending if the attacking network persists). Unless 100% of all economic value goes to their fork, they suffer a loss. Even if they convince 90% of the economic value to move, they lose the 10% which remains on the other fork where their stake is destroyed. Only the non-staking users of the network do not lose any value.

Everything you wrote as quoted above is the opposite of the possible outcome that @anonymint wrote about:

https://medium.com/@shelby_78386/the-caveat-though-is-that-when-the-attacker-can-fork-the-vested-interests-of-some-of-the-users-9340dd037a61

But perhaps the reason you didn't think so, is because you may not have realized the point above about 100% finality is required for 100% objectivity of live observers?

The attacker can profit even in the presence of security deposits.
That was one of the main points of the Medium post.
Apparently there is a great cognitive dissonance in interpretations of the game theory and economics between your thought process and that which is written at the linked Medium blog.
Did the above point about permissioned validators and 100% finality bring your understandings closer together?

Quote from: Ix on June 02, 2018, 12:20:04 PM

If that is the case, the objectively better fork is obvious.

Analogous to network synchrony, censorship can't be objectively proven.
And especially when the censorship is against a few billionaires or such that nobody believes or gives a fsck about.
People have a crab bucket mentality. They love to see some billionaires lose everything.
You seem to not have a very realistic appraisal of human nature and the madness of crowds, or perhaps you just haven't looked it this way before??

Quote from: Ix on June 02, 2018, 12:20:04 PM

But to do so they have to invest in the network itself, unlike with PoW.

Disagree. They gain back what ever they "invested" by shorting one and pumping the other.
Besides they probably bought the hell out of the token when it crashed to 50 satoshis in the crypto winter,
then they pump it up, short, and crash the fscker with an attack.
Or they issued the ICO and bought the ICO from themselves taking 80% of the money at no cost.
Many different sorts of manipulations and schemes.

It's depressing. I want to caution you before you go thinking you have some magic cure.
Many people have thought deeply about these issues for the past 5 years.
You're not the only one.
Although you were probably thinking about non-proof-of-work consensus systems before most of us.
@anonymint was on proof-of-diskspace and then memory-hard proof-of-work ideas for most of 2013 whilst you were already designing Decrits.

Btw @Ix, if you're an excellent programmer and you are interested to collaborate, there's BTC funding and vestment available. But maybe you want to do your own.