← Back to team overview

mimblewimble team mailing list archive

Re: On block rewards

 

The code that checks the block reward of the initial blocks will always
need to be part of the grin codebase (and any other implementations of
grin, should there ever be any).

Because it will always have to be part of the codebase, it seems like extra
complexity for limited benefit.

Interested miners can already mine on testnet to make things are working.

In order to make it easy to switch over, we could add a flag to the client
which takes the genesis and starts mining it on mainnet. This way, miners
can flip their miners over to mainnet once they get the hash of the genesis
block just by adding a flag. This flag could be removed eventually, after a
few releases, since it won't be needed in the long term.

On Sat, Sep 30, 2017 at 7:29 PM Seamus Finnigan <
seamus.finnigan@xxxxxxxxxxxxxx> wrote:

> Hi John,
>
> In retrospect that seems overly complicated. The simplest
> possible schedule, namely a fixed constant block reward,
> probably works well enough. It"s also what Ethereum uses,
> except they started out with about 7 years worth of rewards
> sold in an ICO, judging from https://etherscan.io/chart/ethersupplygrowth
>
>
> The 7 years of rewards, plus the many hints at eventually switching to
> Proof of Stake and reducing rewards, make Ethereum an altogether bad point
> of comparison. Around 72 million coins were distributed via the
> crowdfunding sale and to the foundation/developers. Then another 12 million
> or so per year in their flat emission. That meant in inflation rate for
> that first year was less than 17%, and fell below 15% in the second year.
>
> If a flat block reward is being considered, a better analogy to consider
> would be Bitcoin, Litecoin, or Zcash, each of which had/have flat emissions
> for 4 years. Without Ethereum's premine, they would fit the same scenario.
> Now imagine if after those 4 years (and again 4 years later) there had not
> been block halvings (I will use #s for Bitcoin but the same could be done
> for the others I mentioned). That would mean that today there would be
> around 24.4 million coins available today with an additional 2.6 million or
> so mined each year (10-11% inflation after nearly 9 years), rather than the
> approximately 16.6 million coins and 660k or so new coins per year we
> actually have (~4% inflation today).
>
> Perhaps at the 9-year mark, that's not so bad... 11 percent drops to
> single digits in another year, drops below 5% in another 10 years. But
> those first 9 years are brutal to anyone using the coin as a store of
> value, **unless user and transaction growth match the emission growth**.
> That is what needs to be considered here... the ramifications of a flat
> reward on users of Grin during the nascent years of it's chain. If the
> primary use is pegged confidential assets of other coins, than it's perhaps
> not a big deal. If the primary use is as a currency itself, then it's
> everything. It might be okay, but it deserves discussing.
>
> Of course, if a fee system is implemented that burns fees a la Peercoin
> (as has been suggested by some), then the inflation in the system is
> effectively decreased.
>
> The only deviation I"d like to make from that is a slow start, where
> block rewards are initially zero. This allows everyone to setup and
> test their miners and pools in a relaxed manner, no matter their
> timezone, and offers a unique opportunity to advertise an already
> up-and-running coin where people can still get ready to mine the first
> reward.
>
>
> Testing setups can be just as easily done via a testnet, and it is highly
> unlikely any large miner will test for any prolonged period.
>
> The question is how long the slow start should last. To cover all
> timezones, it should be at least one day. Preferably a few. A week is
> probably more than enough. I propose two options:
>
> 1) At a fixed block height, [rest omitted for brevity...]
>
> 2) As soon as difficulty stabilizes [rest omitted for brevity...]
>
>
> In either case I propose that we try to overestimate the initial
> difficulty
>
>
> A prolonged period of 0-reward blocks would be a mistake. First off, there
> would literally be 0 coins. What do you do with a chain with 0 coins?
> Second, no reasonable miner would spend any longer mining than she has to
> in order to make sure she is set up correctly. Minutes, maybe hours at
> most. If the worry is that she might not have time on day 1, but will on
> day 2, then again - that's a reason for a test net before launch that
> completely mimics the real chain.
>
> The problem here is that the first days or weeks would have basically no
> work behind them, which basically means that come day 1 of the block
> rewards, the chain is still basically 0% secure. But even more problematic,
> any reasonable miner will just wait until the rewards start and then start
> mining, which will accidentally turn the whole thing into a small instamine
> because the difficulty will be too low.
>
> Setting the difficulty overly high is an interesting idea, but only under
> scenario 1, and only if the difficulty adjustment algorithm would not allow
> the difficulty to adjust down to lower-than-expected-demand levels.
> Amending scenario 2, you would want to set the difficulty low, and then
> have some threshold of "X blocks over Y difficulty" as your activation.
> Going high-to-low (depending on the difficulty adjustment algorithm) would
> just further play into the hands of anyone waiting on the sidelines, as not
> mining is the best way to drop the difficulty.
>
> Realistically, if you want to have 0-reward blocks as a startup, the best
> option is probably to set a constant difficulty for that period that is
> based off a best-estimate of the expected hashrate on Day 0, and then a
> target block height that will be the first block with rewards. As initial
> demand is very difficult to estimate (would requiring looking at recent
> launches of other coins), a mechanism could be put in place that will
> adjust the initial difficulty up but not down during that 0-reward
> period... as more miners ramp up to join, if it turns out there are more
> than expected the difficulty can adjust before rewards happen, avoiding an
> instamine.
>
> But I don't really think a 0-reward start is a good idea... there's too
> much guesswork involved, and I'm not sure that it actually achieves any
> tangible benefit. Something on the order of 1 day might be acceptable, but
> any longer is just missing the point. The simplest model for avoiding an
> instamine and allowing people to ramp up has already been played out in a
> number of cryptocurrencies: slowly increase rewards over a 1-2 week period,
> allowing block rewards to somewhat scale with mining demand, and have a
> good testnet available before launch for miners to test out if they would
> like.
>
> Besides, if block rewards are constant in perpetuity, whether a miner
> joins in hour 1 or hour 18 is pretty much a moot point. We're talking well
> under 1% of the first year's coin supply, and under 0.1% of the supply
> after 3 years. Being a few hours late to a slowly and fairly launched coin
> isn't an issue. If someone sees reason to be there from the start, then
> they will make sure they are available and ready to go.
>
> Mischief Managed,
> Seamus Finnigan
>
> Sent with ProtonMail <https://protonmail.com> Secure Email.
>
> --
> Mailing list: https://launchpad.net/~mimblewimble
> Post to     : mimblewimble@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~mimblewimble
> More help   : https://help.launchpad.net/ListHelp
>

Follow ups

References