mimblewimble team mailing list archive

[John Tromp <john.tromp@xxxxxxxxx>]

dear Gellert,

>> If we were to limit ourselves to at most 6 halvings, i.e. a minimum tail
>> reward of REWARD / 64, then with peaking set to occur after 4 years,
>> this tail  won't
>> be reached until 64*4 = 256 years into the future, at which point
>> we're long dead. (In bitcoin it only takes 6*4 = 24 years to reach).
>
> Yeah it is definitely not something we would see in our lifetimes, but if
> mimblewimble ever did survive that long then having a defined tail
> emission after 5-6 epochs may be worth considering. At the same time, like
> you say, it will take hundreds of years for that to ever occur so it may
> not really matter.
>
>> I would be ok with saying that while the formula is beautiful in its
>> simplicity
>> as it stands, we fully expect it to be hardforked at some point in the
>> future into whatever seems appropriate at that time, as long as block
>> reward never exceeds peak REWARD.
>
> Hardforking would hopefully be the last resort for changing the monetary
> policy. This epoch emission scheme looks very reasonable so I am not super
> concerned about that.
>
>> In fact the most radical choice is x=0, which simply keeps the reward
>> at a constant REWARD after the initial ramp-up. And this would be a
>> perfectly sensible choice; probably better than any other value of x
>> stated at the outset.
>
> This was something I was thinking about as well, but the inflation stays
> high for a much longer period of time than the proposed epoch idea which
> may keep volatility too high from a user's perspective making mimblewimble
> wholly unusable as a stable currency for many years. Maybe I am wrong on that though.

If you want to reach minimal inflation sooner rather than later,
then indeed the halvings help a lot. If peaking happens after P years,
then it takes P*16 years to reach an inflation under 2%,
whereas with constant tail reward, it takes P*51 years.
For comparison, with bitcoin it only takes 4*6 years.

>From an inflation limiting approach, 6 halvings suffice to get it well
under control.

> Obviously inflation is to be expected, but it is important it goes down in
> a reasonable manner (hopefully) adhering to two desirable goals. First the
> emission being deemed 'fair' to latecomers because 99% of the hypothetical
> users will come after and if the coin emission overly compensates early
> adopters then they may refuse to adopt mimblewimble due to 'missing out'.

The late-peak, on the order of years, takes care of that.

> And b) that the inflation is both predictable and declines conservatively.
> The epoch scheme meets both conditions much better than Bitcoin's halvings
> every four years, so we are definitely on the right path.

A revised exponential epochs emission proposal looks like:

let epoch i range over blocks [ 2^i , 2^{i+1} )
pick some maximum number of doublings, d, and halvings, h,
and a peak epoch p. let the block reward be
0 for epochs < p-d
REWARD / 2^{p-i} for epoch i in [p-d, p)
REWARD / 2^{min(i-p,h)} for epoch i >= p

regards,
-John