Hey Devs! I have a couple questions as to why certain decisions were taken for the overall architecture.
Why was chosen for a fixed fee (FIFO) on the chain?
Aren’t dynamic fees overal better to allow competitiveness once block-sizes become full and block-space runs out?
What will the plan be once blocksize gets full and the chain clogs? Will it just get a backlog regarding all the TX’s.
Are there future plans for a MEV dAPP which can run on the CCD chain? Or is this not doable?
I understand fast finalization is key, but having a seperate finalization commitee is not great for decentralization, on the other hand we also have other chains who have less ‘‘main validators’’. It just seems like its a bandaid used to fix for finalization + incentives for larger holders to validate the chain. What were the exact trade-offs for deciding this amount/design?
How are the shards validated and managed? I.E. how decentralized and secure are these validators/validations of the shards?
Are there any ideas in the pipeline for implementing roll-ups (OP/ZK) apart from shards? Or do you think shards are enough to scale for world adoption? (seems implausible if even Ethereum does not do it, thats why im asking, or is this feasible due to having a max of 1000 finalization commitee?)
Is the Mempool public or is there no mempool?
How are spam-attacks mitigated if fees drop substantially? Have seen it happen on near/aurora or is this a non-issue.
Thanks in advance, would love to learn more about the decisions!
Here you can find some answers that I got from the science department, I hope these are satisfactory for you. The rest of it will follow a bit later…
Fixed fees helps businesses plan their expenses when using the blockchain. This is also the reason why fees are fixed in EUR and not CCD.
Dynamic fees (like in Ethereum 1.0) have the disadvantage that there is surge pricing (recently there BAYC sold a new NFT and tx fees went up to 1 ETH).
If the capacity of the blockchain is reached transactions will backlog. However, the goal is that this should not happen under normal circumstances. We are therefore developing scaleability solutions that will ensure the required capacity.
The communication complexity of finalization protocol does not scale as well as the underlying Nakamoto-style consensus. We have therefore decided to limit the maximal number of finalizers to at most a 1000. Finalization is thus run by a committee of bakers. Instead of using a complicated election scheme, we use the you-need-1/1000-of-total-stake rule to determine the committee. This should be fine as long as long as most stake is controlled by committee members (this seems a reasonable assumption if you look at the wealth distribution in most countries). We can change the committee election if stake gets more distributed.
Understandable, but wouldnt companies rather have a more secure chain with less trust assumptions and collusion opportunies between a low amount of N finalization commitee? I doubt companies that will use blockchains have a dependancy on a tx-fee that is maybe 1$ or 0.01$ (we will go towards these fees probably within a couple years on main L1s) as they will get more value out of using it for a use-case (is an assumption however, but in my opionion a logical one).
1a. In addition, with what price feeds is the conversion of EUR/NRG/CCD monitored to be paid for the fee? Is it a private oracle or is a decentralized oracle used. + How secure is then the input? (if decentralized which aggregations are used and what are the future plans for this) if centralized, will this always be centralized and isnt the goal of a permisionless system to also have decentralized inputs?
They made them dynamic to have a free fee market. which is efficient (have to admit not always, but often) for users that WANT to pay more (creating tons of demand for block-space).
2a. I have talked with Polynya about this, and FIFO (fixed fee) only makes sense if you are a Layer 3 where users can opt out to a base L2 or L1 with less centralized trust assumptions. Would love your thoughts on the matter regarding FIFO and having a bigger trust assumption with no way to opt out in the system as it is a base Layer 1.
2b. The reason for this thinking, is that for a FIFO system you need to have low censorship in the system, and this inherently means that more trust assumptions are present in the system, no?
Please, Never use the words ‘’the goal is that this should not happen under normal circumstances’’. Almost nowhere in Crypto are normal circumstances. Systems will get pushed to their edges (if succesful) or die off due to no adoption. There (in a long enough timeline) is no middle-ground here. The system SHOULD be able to be resistant to huge loads or attacks. assuming stuff when building is a dangerous minefield.
I have quickly read up on GRANDPA finalization scheme of DOT. It looks almost the same, only difference on the surface is that all nodes participate in GRANDPA for finalization and Concordium has this capped at 1000 (to increase finalization time and thus make a higher throughput which is needed since its fixed fee and FIFO i assume?). Is this assumption of mine correct? If not, what are the exact differences between finalization schemes of the two?
Again, please refrain from using the word should
New question 10: Are there any concrete plans as to what Concordium will do if adoption rises heavily and there starts to built up a backlog in the system? What resources do you guys have to fix that? Can finalization time be increased by reducing number of finalization comitee, but thus increasing your trust assumption of not colluding finalization members, or do you need to use other tech such as Danksharding/Rollups. If so, What plans are in the pipeline currently for that? Increase block-size maybe?
I understand that the project is in its infancy and I do not know everything either. So would love to see where my thoughts are wrong, but it is up to you guys to come with counter-arguments. In addition, if you guys do not have the answer yet thats totally fine too, rather have an answer as: ‘’we do not know yet’’, then making a less solid argument that falls off later!
Again, I really appreciate the time and energy you guys put into in answering all my questions!
Here you can read more answers from the Science team:
Fixed transaction fees do not preclude the blockchain from being secure and trustless. We can have both.
1a. We currently query the exchange rates from coinmarketcap, coingecko and livecoinwatch, and take the median over the last 60 minutes. This is updated every 30 minutes.
No, fixed fees do not require more trust assumptions than variable fees. The blockchain works the same under the same trust assumptions in both cases.
It should not happen in good weather conditions (most reasonable chains run most the time in those conditions). We are not too worried about spam attacks (we can adapt fees using governance operations which have priority over all other transactions). The security of the blockchain is not affected by this.
The speed of finalization is not a limiting factor in the blockchain, so changing that will not increase the throughput. We have many projects in the pipeline such as a faster consensus protocol, sharding, zk-rollups. But we don’t know yet what we will push out of the pipeline first when needed, so we can’t give you more details at this point.
If you have more questions like this, I would like to recommend our telegram /Telegram: Contact @concordium_official), where you can have a more sophisticated conversation about these topics.
