Topic: NA - page 323. (Read 893613 times)

Click on the pictures for a larger view.

Overall testnet chain difficulty chart at 20% step limit:



The first 2 jumps are from baseline to 5 times the network hashrate.  After that, the jumps are 10X increases, and the highest peak is when we toggled a 15X increase for giggles.  In the following pictures you will see the difficulty values and the steps up and down instead of instant spikes and crashes.  Notice that each time the hashrate was removed to go back to baseline, it never dives below the baseline.  It tapers back, and when the block is found, it stabilizes instantly.  This is something DGW3 isn't doing properly.

5X increase in hashrate:



15X increase in hashrate:



10X increase in hashrate:




It's getting late here, and the wife wants to watch The Interview, so I'm taking off.  I'll be back in the morning to answer questions and provide any needed info.

-Fuse

While I can understand the skepticism, I will assure you that my team is dedicated to helping this coin however we can.  So we decided that regardless of what was happening, we would test DIGI anyway on a testnet.  I've got a block explorer with the data on it, so I'm putting it together now.  I might install my data analysis software from work here so I can do a bit more than excel can.  So it might take me a little bit, so hang tight.

As far as Terk liking DIGI, I'm not sure he "likes" it per-say.  In my PM convo with him, I feel like he was at least a little sincere when he said it was the right move.  At that point, I don't think he was the full-bore asshat he ended up being now.  My initial thoughts were for DIGI, way before we even knew it was clever that was messing us up.  I've seen proper DIGI implementations with coins that had 2-3 times the hashrate we have had thus far.  The instant the DIGI went live, MPs dropped off and mining was stable.  Sure, clever will still mine NLG, but they won't pull 20 blocks and leave us with a sky-high block that takes an hour to solve.  Maybe they make off with 1 or 2 before they are left unprofitable.  At that point, it might not even be worth it to them to mine, as the daily profit is significantly less.

All that being said though, we haven't proved the reliability of the simulator results.  While I am 100% behind it, how do we know that the results we'll see with it will be the same as with a testnet?  It is in fact a first version that has really only been tested by a handful of people with a small subset of variable conditions.  If we treat this like any other scientific/mathematical research, it needs to be peer-reviewed, tested and substantiated.  I'm not saying that we absolutely have to do that.  I'm just saying we need to know it's results are as reliable as an actual testnet.  To fully stand behind it as the sole decision making tool is just as much of a leap of blind faith as trusting me.  We still need large wave simulation and we need to verify that data that we find is similar to what we would find on a testnet/live chain.  It is a simulator, not actual mining... there could be variables that just won't present themselves with it.  Again though, I'm all about it's development and couldn't be happier that it was made.  I just don't think we can wait to get it sorted before we work on the algo.

AGAIN... not knocking the simulator.  I just want to make sure we are using the right decision making tools, and using them in a timely manner.

Just my 2 cents.

-Fuse

So it would be beneficial to make a working DIGI-plugin for GJ's simulator and get accurate results and tweak it to perfection. We can look for GO devs to do it for you Fuse, because it would be handy having one or two extra set of hands with this specialty code anyway.

We could probably get some devmine to pay a dev for the work. Either straight converted into USD or EUR or if the dev wants a stake in NLG that is obviously better.

I know a few places we can post a job with general info and pick a freelancer. I agree that we should know what algo GJ is working on getting functional in the simulator so we don't waste resources. That way we could efficiently move forward and test multiple algo's and choose and tweak the right one. So details on what is being worked on instead of in-depth details or deadlines would be much appreciated. That way we don't trip over each other's feet.

As long as we are not willing to reject extrem short block times (a time that is clear below a reasonable 'lucky block'), there is always a nice window for large hash powers until an algo can catch up. The size of the window depends on how fast the algo can react (and important too that it does not overreact).

Clever likes KGW and (standard) DIGI because Terk has tuned his predicting algo for those and exactly knows when to jump in and out. Time has proven, DGW3 isn't a problem for him either. So we need an diff algo that can show us, it will react better than what is out there at this moment.

The simulator is needed because noone really wants to test one algo after the other on the live chain.

While I have my doubts, I will applaud Fuse and his team if his DIGImod will show the promised results in the simulator.

Fuse: Just saw your post and want to say I am very interested to see your test results in the morning.

No, you're on point, mate.  I never released our results publicly.  And I will be completely transparent in saying that our tests were done on a smaller scale, but with proportionate hashrates.  We got baselines, then threw 10x the hashrate at the chain.

Reaction times were very quick, and extremely effective.  Difficulty ramped up fast, and dropped off in a controlled fashion back to baselines, instead of plummeting into insta-mine territory.

Instead of the traditional 10% limit DIGI uses for per block difficulty jumps, we upped it to 15, 20, and 25%, testing each time.  The increase was to account for the fact that DIGI is usually implemented with coins that have a block time closer to 1 minute.  That means that there are 2.5 times more retargets per hour/day/etc with a 1 minute block.  Our block time plays against us here, and I'm thinking it might be part of the reason DGW3 isn't working as intended.  So we found the magic number to be either 20% or 25% for DIGI.  You could shorten the block time(and rewards to match), but that's a major change.  I think the change would help, but I'm not going to push for it at this time.  If you don't increase the number of retargets(shorter block time), you need to increase the difficulty limits.  So the results I will post in a bit is for the 20% limit.  25% was great, but it might be too aggressive IMO.

As far as another change down the road... well we can't predict what will happen down the road.  What's to say that quantum computing doesn't hit next month and we are in a world of hurt regardless of what we do.  I will tell you that coins that implemented DIGI well over 4 to 5 to 6 months ago are still solid with it.  It the algo Terk himself recommended, and the one I pushed for for almost 4 months now.  There's a reason behind that- it works.

I'm holding the baby right now, so this reply took me quite a long time to write.  I would post the stats right now, but this is proving difficult lol.  Give me a little bit and I will post charts and some explanations.

-Fuse

As a long term Guldencoin investor I understand your frustration.  When implementing something new like a cryptocurrency, it is important to work together as a team for timely coin development.  If miners, investors and the public see a better coin they will move on.

Fuse, I hope that you and your team can work with GJ on an algo action plan.  Time is important!

@Fuse do you have good test results for DIGI if so, how good are they? Will the change most definitely make mining fairer to all? So fast and precise retargeting without a window for any swings worth jumping on?

If so, do you have reports for these results? I take your word for it that you think DIGI is the best option, but the simulator was made because of its direct and precise measurements. I want to see a solution as well, but one based on hardcore maths and stats. Not because something was tested and worked okay. Are you 150% sure DIGI will work perfectly and why? Did you make changes so it would suit NLG or is it standard DIGI?

I am not dissing your claims here, but just want to know the numbers and how effective it is to cure our mining. Multi's will not stop mining NLG and we don't want them to, what we want is fair distribution and for everyone to be able to mine and use NLG normally. Can we do that with DIGI or would we need another change in 6-12 months?

Sorry if you explained this all already, but I do not recall reading it.

Cheers!

I used the word 'agenda' because I really don't know, and I think this counts for everyone, when, what or how the devs will make the or even a change.
It's just fuzzy if you ask me. While time passes by trust slowly leaves...

I myself, don't have the skills or knowhow to solve this problem, so I can only give my short vision about this matter and so, if you ask me, I like a modified DIGI more then nothing right now. Hope /GJ is working om something better but I just don't know. Quite an impasse if you ask me because I still have a huge trust in the dev team. Hope to discuss this with them soon too!

Anyway, great to hear we have a solution which can be implement quick. Think people with GO knowledge are scarce? I took a quick look and I think JAVA/C people can learn GO quick and easy? But, someone also need to have some domain knowledge too? That's the main problem isn't it?

Easiest action is implementing DIGI.  However, /GJ made it clear that he wants to see a GO port of the DIGI algo for the simulator.  But besides the DIGI algo, we need the additional wave simulations that /GJ needs to implement for that testing to be completed.  Testnet work was done by my team, and we're confident it would be the right move.  But we're not the dev team, and our opinions aren't the only ones that matter.  That's why I've been asking for the dev team to come and discuss an action plan for the algo change.  I don't need an ETA for when the change will happen, I just need to know that we are working towards a change.

I think you're 100% correct though in people losing interest/patience because of the delay with the algo.  A member of my team pulled his hashrate from the network because of it.  It's not a substantial amount, but it's not little either.  Another is considering it.  You have to think that there are a handful of people like them that have done so as well.  I'm not saying we need to rush, but we need to not be sitting here waiting for dedicated miners to jump ship and CM to take 80% of the network.  We need to formulate a plan, and the devs need to be involved in it.

Just to let you know though, if I was given a block number when the change would go live, I could post the modified DIGI code git pull request right now.  It's not a difficult code change.

-Fuse

After the raise in price Clevermining start 'using' the opportunity again.
So there is where price manipulation, besides dedicate mining, can make the shortterm difference in network stability.
Seems like some dedicated miners stopped or lowered their hashrate too...

But for sure, we need some chance now, we wait far too long, some people lost their interest/patience and walked away from Guldencoin.
Most (or at least many) projects are postponed.... etc etc

Reading back learns that the Guldencoin team works 100% at new algo. They can use some help too, don't know if someone offered some in the meantime?

Also,

@Fuse,

you mentioned DigiShield. Did /GeertJohan respond to this or is there another agenda? What is the easiest action to stop this CM terror?

When I looked this morning they where at 72% so an algorithm action plan would be something we really can use.

WIN!!!  120 Hours of free 20+ MH/s Scrypt mining. 
Only signup at:
https://cryptocorner.nl/activity/

For the very first 250 members, we'll take out one winner!

Any thoughts on formulating a algo action plan?  I'll give you a few reasons why we should:



-Fuse

Oh well, again, too late..

Lol, prolly Fuse can arrange something about that

To clarify, time now is 11:12 PST.  So people a little over 12 hours to take advantage of the deal.

-Fuse

Keep in mind it's 7:35 in the evening right now in the Netherlands, today is almost over. Maybe a bit confusing 

oh well i bought 1 day too early haha

Merry Christmas, everyone!

20% discount on the coin store today only in celebration of Christmas:

https://bitcointalksearch.org/topic/m.9943116

-Fuse

@strataghyst  thanks! have a good x-mas