Thanks MegaMike. I checked the links on Twitter. The most informative, but not really insightful, I could find was http://forum.megacoin.in/index.php?topic=2791.msg9395#msg9395, which lists various Kimoto Grawity Well configurations. In case of MegaCoin it is:
[quote]
Megacoin
Date Added: September 13, 2013 (~3.5 months in)
http://megacoin.co.nz/
EventHorizonDeviation = 1 + (0.7084 * pow((double(PastBlocksMass)/double(144)), -1.228));
BlocksTargetSpacing = 2.5 * 60; // 2.5 minutes
PastSecondsMin = TimeDaySeconds * 0.25;
PastSecondsMax = TimeDaySeconds * 7;
[/quote]
This leaves me clueless since there are no explanations of what PastBlockMass is, how it is calculated etc., what do the numbers in EventHorizonDeviation actually mean… I am not even convinced that I am looking at the real thing as it looks too simple to work as intended.
Briefly inspecting MegaCoin difficulty adjustments it can be concluded that it works OK. And so does the diff adjustment in the case of the Feathercoin, just in a bit less granular way.
http://www.coinwarz.com/difficulty-charts/megacoin-difficulty-chart
http://www.coinwarz.com/difficulty-charts/feathercoin-difficulty-chart
I don’t think that at this point â€Å"a bit less granular way†of diff adjustment is a problem. On the other hand it may become a problem, even a major one, if multi-pools become very strong and hit us with terra hashes at one moment, and leave after the difficulty rises above certain level. With a substantial drop in hash rate even solving a single block could be a major headache.
Would I vote for implementing Kimotos’ Gravity Well? I’d like to understand how it works first, and see how it fits the pessimistic scenario of rapid and high-magnitude hash-rate changes.
Maybe we should be worried about the scrypt asic dominated future, and see if we can come up with a suitable algorithm on our own (not simply implementing what others have already done), perhaps designing an algorithm that is very sensitive to increased network hash rate (so that we only waste a couple of blocks when a multi-pool hits us with terra hashes), and on the other hand reacts super quickly when multi-pool terra hashes are gone (perhaps using a failsafe if a block is not solved within a certain time frame*)
* This could provide incentive to a powerful miner to try to solve a couple of blocks (or even e single block) after the failsafe kicks in.