Topic: POW + POS idea.. (Read 522 times)

I'm currently looking at a potential solution to POS - replacing it completely with proof of burn... if I get past all the design hurdles I've got atm, I'll post a call for scrutiny on here

Yep..

As I see it you would run the POS chain first, on top. Just like a normal POS chain.

And then the POW chain would be playing catchup, time-stamping the POS chain blocks.

You both(POW & POS miner) get 'paid' once the block has been POW-Stamped.

I haven't thought it through enough to get an algorithm, but something like a 10 sec POS chain (sprinkle of GHOST), with the POW chain maybe operating 1 minute behind..

So, as you say, there would be 1 minute where the miners could 'play' around, for free.

I'm sure there would many advantages to such a system, but there is certainly a large amount of complexity being added..

Nothing at stake is the subset of attacks which become possible in a system where block production has no cost. It has nothing to do with the process of acquiring stake.

You can run through each of the currently understood attacks one by one and apply this bandaid or that bandaid, but really you have to question the wisdom of a system which does not solve the byzantine problem in the face of sybil attack.

If producing a block has a constant probability proportional to historical stake, then so does a double spend.

Not 'fixed' but reduced in scope to the interval between POW blocks.

By "Nothing at stake" you mean the attack described in this link http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2393940 [first google result]?
If so, the attack requires 51% of forging stake, which would pump the price to very high amounts. Moreover, the attacker is stuck with 51% of forging stake, which in NXT would [after bubble] mean burning tens of millions of dollars. That sounds like a much-at-stake attack to me.
Or do you mean some different N@S attack? If so please provide a link.

Ad. 1)
Right now when someone synchronizes, he downloads the chain with the most cumulative hit. [cumulative hit == most POW work put into it]
To be able to fake cumulative hit you'd need to have > 50% of forging stake at any point in time since the last hard fork. An attacker would be burning tens of millions of $.
Ad. 2)
You cannot create blocks into the future and present them as valid to unsuspecting client. A new client checks the timestamps, and if they're invalid rejects the blocks.
[Assuming I understood your description correctly. If not, please elaborate.]
Ad. 3)
How would you exactly implement this system?
Depending on the variant PoW miners may have hands tied so strongly that they have no defense against 51% stake attacker, or the attacker could attack the network with 10% stake & 51% PoW.
"The miners can't censor or be coerced to." - If miners can select which blocks get into the blockchain, they can censor. What prevents the government from building ASIC's? What prevents the governments from giving 51% of miners $ for their efforts?

1) Even if PoW would add security [it woud decrease it if implemented incorrectly], it'd rise the costs of keeping the currency considerably. Compare inflation of PoW currencies [8.2% currently] with for example NXT [0% inflation, financed 100% from fees] and you'll see why it'd be a system with considerable drawbacks.

Well..

POS has various attacks that are performed because of the NaS (Nothing At Stake) issue. Yes I am fully aware of the current pseudo-fixes and that's fine. But with this, no fix is required at all.

Just as POW has some issues with mining centralisation.

With this :

1) POS Long Range Attacks. The POW time-stamping would remove the 'Weak Subjectivity' that POS suffers from. So no need to ask a friend for a checkpoint / hash when you log in for the first time. It's there just like with bitcoin, as the chain with the most POW work put into it wins.

2) POS Short range attacks. You can't simulate infinite futures as the POW must also be included.

3) POW mining centralisation would seem 'less' important. The miners can't censor or be coerced to. They are JUST the security of the system. The POS block creation has a much larger user base. More users than miners.

So in this system, you would have the advantages of POS systems in terms of WHO ACTUALLY DECIDES WHAT HAPPENS, and the advantages of POW systems in terms of SECURITY.

Win win.

As for time stamping, NXT does it pretty well without PoW. I don't see any "costless attacks on PoS". Can you elaborate what advantages your system would bring over existing PoS implementations? Can you describe your algorithm in pseudocode or just in more detail?
Right now I don't see any advantages of mixed mining that you describe.

ADDENDUM :

..

Obviously, having the chain timestamped with POW, means that all the costless attacks on POS are now fixed. 

POS is simply used to pick the creator of the next block. Who can then create it. (And NO you can't run a costless simulation, the POW is real..)

There is a payout to both the POW and POS miner in the block. Though not necessarily equal.

..

END of ADDENDUM

 

Lot's of talk about Consensus mechanisms on the boards ATM.. Get's the old imagination going..
..

Miners are 'really' just time-stampers.

That's the Battle that wages between them.

Who get's to decide the TIME.

And they get paid 'well' for that. And that's a nice, neat, clean relationship. 

..

In current block chain systems the miners ALSO get to decide the Content, Which events and the Sum of the Events , that occurred during the last slice of time.

Why not change that so that Miners ONLY worry about the Time. Something they are good at.

And the actual Content, The Blocks, are created by a different group. Say a POS mining group (quite a few options here..) .

So POS Miners are busy running full nodes, NO hashing, and create blocks with TXNS in them. Sign them in some fashion.

POW Miners are busy running full nodes, FULL HASHING, and crunching mega-hash functions, to solve the current POS Block.

I can't see that the partial-centralisation of the HASH mining would be such an issue then.

This would distribute the business of creating the 'important' Content 'bit', to a far wider audience.

?