More than that though. It allows for lower orphan rates with faster blocks. Also limits benefits of selfish mining.
Imo there is still reason to explore tailstorm, but faster blocks and abstracting block time is the first step.
2 Likes
No, orphans (even though they get merged later) still happen at same rate. The trick is in reducing and socializing reward losses, so miners’ share of revenue still matches their share of hashrate.
This potentially achieves the same, in a much simpler way: A Deterministic Tiebreaker for Bitcoin Block Selection: Enhancing Fairness and Convergence
That table makes it hard to compare things.
I take it the 6.67% (for T=75) refers to orphan rate of summary blocks, but with k=15 the inner blocks are at 5s and should see a huge uncle rate.
With Tailstorm:
- Each merge redistributes payout from miners who’d have won the orphan races to miners who’d have lost them.
- Each merge contributes to PoW accumulation.
The simpler alternative also redistributes payout from would-be-winner to would-be-loser, but doesn’t include the losing block, so one block’s work gets scrapped whenever there’s an orphan race.
What’s the objective? To make mining fair (by fair I mean that %hashrate matches %revenue) or to reduce scrap work? It can be fair while still producing scrap.
I believe this is for sub-blocks, not summary blocks:
“The orphan rate, that is subblocks orphaned per subblocks included, is bounded by [above formula]”
Ah yes, sub-blocks are what’s getting orphaned (especially in the implementation where summary is a 0-PoW deterministic block rather than adding 1/K too), but orphan races all happen at summary block time interval. I believe reduction in orphan rate is just because of the full download assumption. With Tailstorm: a node will have already downloaded K-1 sub-blocks, and if 2 subs are racing to get picked as last sub to complete the summary - a node will only have to download 1/K of data. So: with compact block relay (or high enough bandwidth) this difference in orphan rates should be negligible.
Difference is in cost of these orphans, with Tailstorm the orphan will cost only 1/K reward, which makes them less impactful to mining fairness because they’ll reduce miners’ revenue only by 0.44% (relative to K blocks worth of rewards) rather than 6.67%.
This is all ignoring uncle rate.
(Uncle+orphan) rate for 10s subblocks should be the same as orphan rate for 10s blocks.
Can you TLDR me, what is going on here? How can I best contribute to the discussion? Is it productive to discuss the merits of Tailstorm here? I think we have done that excessively in another thread.
@bitcoincashautist should consider adding the benefit for node operators since each block will require less memory and CPU cycles.
Absolutely. Just high level!
In terms of the above couple comments (two brief points below):
- Could the deterministic tiebreaker
be a way to assume most of Tailstorm in a simpler matter?
2. Would uncle+orphan rate on Tailstorm for x second blocks be the same roughly as orphan rate for x second blocks, and very high level what the actual end difference would be if any.
Sorry yeah I know largely discussed but in this context just curious to hear from your perspective!
1 Like
Deterministic Tie Breaker. My understanding is that whenever honest miners see two competing chains on the same height, they will deterministically select one of the chains as preferred chain. This is in contrast to the current solution where miners prefer the chain first received. This change ensures that all honest miners agree on the tip of the chain, in cases where it is ambiguous.
In a benign network, where everybody acts honestly, this would reduce unfairness. But as a pool operator I would certainly modify the software to prefer my own block instead of discarding it. If all pool operators think like me, we’re back at square one regarding fairness.
Regarding selfish mining: I assume you’re familiar with the attack models of Sapirshtein/Sompolinsky/Zohar or Bar-Zur/Eyal/Tamar. They give the attacker some means to reorder messages: if one block of height n shows up, they can send their own block of height n (if they have any) with the MATCH action. Fraction \gamma of the defenders will continue mining on the attackers block. If you now add your hash-based tie breaker, this reordering will work in 50% of the cases and should be at least as bad as assuming \gamma = 0.5 in the other model. I say at least as bad because the attacker knows in advance whether the MATCH will work and probably can adjust his strategy based on that.
Uncle+Orphan rate. I’m not sure what you mean with uncles, as Tailstorm does not have any. It has trees of k subblocks between each pair of summary blocks. It discounts rewards based on the height of this tree. Probably you mean, if the subblock tree has height k - i, there are i uncles on this tree? Then the uncle+orphan rate in tailstorm with subblock interval n seconds should be about the same as the orphan rate in BCH with n seconds block interval. Assuming propagation times of Tailstorm subblocks and BCH blocks are the same.
1 Like
It’s assuming you’d do this, so dominant pool hash-rate wise still has its advantages. However, it would use the tie-breaker on blocks received from others, so pools with higher ping would still have a chance of winning races.
Wouldn’t expected ROI be negative? E.g. you had a valid block with no competition and you withheld it then in doing so you allowed someone else to mine a block that will flip yours with 50:50 chance.
What if we don’t give attackers the time? Independent researcher zawy was suggesting to enforce block timestamps to be monotonic (each block must advance at least by 1 second) AND have nodes have very strict future time limit (e.g. max. 6 seconds ahead of their local clock).
It has whole uncle branches, we could call it branch rate or something. Anyway, thanks for confirming that the branch rate is the same no matter what you do on a branching event (orphan or merge).
All these schemes rely on changing the impact of branching. We can’t remove branching at the lowest unit of PoW.
Yes, that’s right.
The mentioned papers make it clear that the ROI is positive for strong miners and \gamma = 0.5. It’s even positive for \gamma = 0 (no reordering possbile). I argue that your change achieves something similar to \gamma = 0.5. Whether this an improvement or not depends on whether you find \gamma = 0.5 a realistic network assumption.
Well, withholding gives the attacker more time to mine the next block. Your flip scenario applies if the attacker is unlucky and cannot mine a block in time. But if he is lucky, he now has a private chain of height n+1 while the defenders are still indecisive about height n. With the deterministic tie breaker they are decisive about height n, but this does not help with respect to the n+1 chain.
2 Likes
I support this. It does not add any barriers for implementing proper economic security through Avalanche down the line. An argument that seems to be missing in this debate, is that faster blocks increase censorship resistance since it’s easier for solo miners and small pools to find a block. As long as orphan rate is reasonable and you don’t centralize because of this.
2 Likes
Unsure if asking for my thoughts on something specific in this thread or the CHIP in general.
I do not want to rehash the discussion about the merits of tailstorm so i will focus on the issues i have with the implementation of the tick system outlined in the CHIP. To me it seems his chip seems like a giant trade-off between getting some features of tailstorm and having a tick system spread throughout the codebase vs implementing tailstorm itself and not needing the tick system.
I understand tailstorm itself is a decent amount of work to implement but it does not touch the external systems nearly as much as this tick system. A lot of the tailstorm changes are internal to the tailstorm protocol and would only be preset in the tailstorm section of the code.
As noted in the technical description here, the tick system changes the subsidy schedule, DAA, ABLA, timekeeping in both script and the tx level, coinbase comments, sigchecks, blockheader rate, and rpcs. Tailstorm does not change most of those things in any significant way.
- DAA, ABLA, timekeeping, sigchecks, and block header rates - All of these are unchanged in tailstorm. They would only update every ~10 minutes when new summary blocks are published. Same as they do now for bitcoin blocks.
- coinbase comments - i do not think these would change at all either but i would have to check the bch tailstorm testnet code to be sure.
- subsidy schedule - This would remain the same at the summary block level in terms of blocks mined. no shortfall that would cause it to end sooner. The total amount paid out might be slightly less than the 21M target, but the last block that pays a subsidy is the same.
- rpcs - yeah, additional rpcs would be needed.
- As for future proofing… Because there is no need to store all full subblocks after a summary block is formed, changing K should always be a backwards compatible change keeping the calculations that involve K relatively clean. add in some if statements relevant to the fork, post fork remove the if statements and only use the new value. eliminating the need for annoying calculations like this one
tick = H_a * 600 + (H_b - H_a) * 60 + (height - H_b) * 30every time there is a HF to tweak the tick rate
Idk. i find it preferable to keep the changes contained. I am not particularly a fan of the tick system for that reason. it touches too much.
2 Likes
Hey thanks for having a look at the CHIP! (note my orphan rate analysis is off a little, need to correct it)
What makes Tailstorm Tailstorm? Is it the uncle branch merging / reward system, or the faster-blocks-as-a-soft-fork aspect of it?
You could retrofit Tailstorm merging to this tick system, and we could just have faster blocks as “subchains” without the merging system. Basically we have 4 options:
| Uncle merging | No uncle merging | |
|---|---|---|
| Target time change | Outer Tailstorm | Faster Blockchain |
| Blocks subdivision | Tailstorm | Subchains |
Talked about “outer” Tailstorm here.
So your preference is with the subdivision approach, got it! Uncle merging only affects consequences of branching rate. The inner/outer approach affects impact on downstream software and affected consensus rules. Let’s just assume we don’t want merging and focus on this aspect. Call it subchains.
I already made a comparison table here, but you made me realize there’s more to consider.
Does this mean that with original Tailstorm you don’t get the benefits of increased time granularity in contracts?
Those wanting to consider subconfs still need to switch to some new rpcs. Difference is opt-in vs breaking. Thankfully we can update RPCs ahead of faster blocks HF and encourage switching beforehand so the HF goes smooth. Example:
There’s another way to eliminate it: turn the block time into a state associated with each block like ABLA’s internal state. That would make it possible to have variable block time (like Kaspa and Nervos have), but I think there’s no need for that and has many unknowns on game mechanics.
Anyway, it’s a simple calculation easy to extend a bunch of times, we don’t expect to make more than a few adjustments.
This is an excellent point!
1 Like