Topic: [ANN] Catcoin - 0.9.1.1 - Old thread. Locked. Please use 0.9.2 thread. - page 7. (Read 131044 times)

I would submit and make available the Penta algorithm developed in C#, but we would need a volunteer programmer to translate it to c++ and provide a couple of conversion functions to translate native difficulty representation to the floating point representation and back, to bring the code to the state where it can be tested on a testnet. The code is already testable in the simulator at catcoin-algo.org but it is understood and agreed that code must be converted to C++ and its performance must be validated in the testnet environment before it can be a candidate for implementation.

Overnight, testnet was changed to run Digishield with a 1-block retarget.  We'll see how it behaves with very low hash rates, as well as when assaulted.

http://testnet.geekhash.org/

We continue to seek out and explore various exploits as well.  We intend to select the difficulty adjustment method that provides the most even delivery of blocks in as broad a hash rate environment as possible, and one that is as immune to time warp and other exploits as possible.

Thoughts or suggestions?  Bring them!

Andy

Testnet...

Still working with DGW3.  The last round of experiments did improve the reaction speed of the algo when the hash rate increases by more than 30x, and the difficulty 'can' come down fairly quickly.  The steady hash rate performance isn't as smooth, though, and exhibits slow cycles that result in long blocks on top and difficulty dips in between.  A hash influx of about an hour is enough to leave us high and dry.  It appears that DGW3 can provide some resistance to massive difficulty changes, but we have to pay for that with fast blocks early, and the probability of some very long blocks on the other side of the 'attack'.

I think it's time to put DigiShield up and abuse it for a while.

If anyone's got other ideas, this would be a great time to post them.  Thanks in advance.

Andy

http://cat.e-pool.net:9993/static/

Enjoy e-pool mining!

The new primary seed node, catcoinwallets.com, and the block explorer are running on their new server tonight.  DNS propagation should be complete by now.  There's a message on the old server - if you receive it, please post the message.

Testnet was reset a bit ago as well.  The algo's still DGW3, but we changed some of the settings a bit.  We're hoping for smoother performance in steady state and a bit less overshoot.

That's all I have for now.

Andy

DGW3 performs better when the block look-back is trimmed to 12 from the default 24 blocks.  It's not quite as smooth during low diff periods, but responds much better when a 40x hash rate increase happens.  The 36 block average at the top (not including the hour-plus top block) was 6:30.  I expect the block average to exceed 10 minutes on the 'down hill' side when that happens.  Right now, though, I'll leave my 30 MH/s and see how the algo acts now that it's adjusted to the hash increase.  (So much for expectations - things are coming down much faster than I thought they would...   )








The bad news in all of this is that when my 33 MH/s is on testnet, it's not on mainnet.  Expect mainnet blocks to slow for a bit...

Andy

We've just adjusted some settings to see if we can speed DGW up a bit...

Thanks for checking in!  What difficulty adjustment process do you like?
Andy

By the way - this is Digishield in the real world with more hash.  Look familiar?

That is the only problem with the DGW.

I'm hitting testnet right now with 33 MH/s after about 8 hours of less than 1 MH/s.  DGW3 is very slow.  It takes more than 20 very fast blocks before the difficulty starts to rise.  It also takes way too long for the diff to fall after the heavy hash leaves.  It's very smooth at low diff - no sign of dips - and handles gentle ups and downs beautifully.  But an attacking pool can hit us for 5 minutes and leave us hanging for hours.

I hate to say it, but of DGW, digishield, or our current PID, PIDs in third, DGW in 2nd, and Digi's first.

I still want to try the Phoenixcoin multiple moving average method, and want to find a way to change Digishield's 1 block retarget to 3 or 5.  We'll see.

http://testnet.geekhash.org/

Andy

We now have a Catcoin daemon with the Dark Gravity Wave 3 diff algo installed.  Blak's got it for testnet.

Let the fun continue.

Andy

DGW3 is on testnet tonight.  So far it's pretty smooth.  A number of coins have moved from KGW to DGW to Digishield.  I guess now we'll know why.

Edit...   Just hit the network with my 30MH/s.  We had a pair of Gridseed 5-chip miners on, for about 700 KH/s at the miner, so this is a good hash jump, but nothing like some of what we've seen on mainnet.  The algo easily adjusted to 10 minute blocks - so did a very good and smooth job in steady conditions.  So far, the algo's pretty slow to respond to a hash increase.  Got about 15 blocks in th e5-15 second range and two blocks at a minute-15 and a minute 50 seconds.  It'll be interesting to see how the algo respons on the 'downhill' side...

FYI - don't consider these rounds to be formal testing.  I'm basically just checking my code to make sure the algo's work well enough test. 

You've already written the code - and you kept trying to get us to run that as well.  The rest of us were trying to figure out how to fix the actual problem.  (Come to think of it, why weren't you using your 1337 skillz to fix the algo instead of trying to harpoon whales?)  Bzzzt - thanks for (not) playing.

Great.  Let's just jump from "Gee, my last takeover attempts didn't work...maybe I can find some suckers to try this one.  MINIONS!  I need minions to run my code!"   Bzzt - thanks for playing.

No, "he who hozes" - nobody is "pissed off" because you "go and test something" - we were damn angry because you kept trying to push your latest exploit and We All Said NO - and you did it anyway.  Do not even try revisionist history on top of the other layers of BS.   Bzzzt - thanks for playing.

More lies with a side of FUD.  Nicely done, hozer, nicely done.  Except you knew damn well Cat wasn't an orphan, you knew it wasn't up for a take-over (though God knows you kept trying).  And you also knew full well that we were putting time and money into this project - you and I spent enough hours on the telephone talking about that, even if you weren't paying attention in IRC or via email.  What you didn't know is that I've been learning C++.

Bzzzt - thanks for playing.

Cat pictures...couldn't hurt...

Some info on testnet, and a request for additions to the test plan.

First:  The code running on testnet can be found here:  https://github.com/SlimePuppy/CatcoinTestnetDigi

Second: Most coins have special testnet difficulty rules so that long blocks result in testnet mining a minimum-difficulty block to get things moving again.  We have disabled this function.

Third:  The current release code has Hozer's minimum block time kludge.  This was installed temporarily as a way to make PID's low-difficulty periods 'less bad' after it started to gallop so badly on mainnet.  This has all been disabled on or removed from the testnet code.  (If anyone finds remnants of this still active in the code, please yell - we want it ALL gone.  Thanks!)

Fourth:  We're trying to duplicate mainnet as closely as possible on testnet.  Same code, same levels of hash input, etc.  We found that testnet results do reflect mainnet performance, and that things we see on mainnet are able to be duplicated on testnet.


So far, we've seen that Digishield returns an approximately 13 minute block time when the network has very little hash rate.  It doesn't 'lock in' on any specific block length but isn't expected to because of at least in part the random nature of the 'block solving lottery' and the way hash rate fluctuates in the real world.  It handles large hash inputs easily and quickly with no sign of insta-mining blocks.  On the 'downhill' side, there's at least one long block (the current block is about an hour long, after dropping from ~33 MH/s to ~3 MH/s).  So far there's no sign of cyclic responses** or massive difficulty drops or abnormally low difficulty periods that would bring us the types of short duration hits we've been seeing on livenet since Cat was born.

So...what kinds of tests should we run against the algos as we work to implement one?   What tests must the new algo pass?

Thanks folks!
Andy

**  Oh well...yes Virginia, we have cyclic responses.  The algo's not working very smoothly during extremely low hash periods, and it leaves us high and dry when we're hit hard.  I guess the good part of this is that because there aren't any super-low dips, it won't be profitable for switch pools to camp us long enough to get the 10 or so blocks they're getting now.  This is better than our current PID system, but hopefully we can do better.

As promised:  Testnet's live with Digishield   


Testnet:  http://testnet.geekhash.org/
IRC: #catcoin-testnet

Right now it's quiet with two 5-chip Gridseeds and a pair of GAW Fury for about 3 MH/s at the 'pool.'  I'll pull my Furys in a bit and put them back on mainnet as we've got a hell of a block to break.  Maybe on the weekend we'll call for some coordinated scrypt hash to see how it handles a hard hit.

Andy

This evening I've forwarded code to Blak for Testnet.  It's CAT with the DigiShield difficulty adjust.  I'm almost done integrating DGW3 into a catcoind as well.

Now that we're not wasting time with BS whale hunts we can get some much overdue work done.

Here's to a long overdue stable Cat.

Andy

Btw, if any of you want to make money, figure out what the hell this means

http://www.businessinsider.com/afp-ecb-jan-22-meeting-to-examine-launching-sovereign-debt-purchase-2015-1

http://www.forbes.com/sites/chriswright/2015/01/15/why-the-swiss-franc-shot-up-30-in-a-morning/

http://www.wsj.com/articles/swiss-franc-move-cripples-currency-brokers-1421371654



Bailout 2.0 anyone?

First, thank you for finally saying something. You're the only person that posts here that has actually taken the time to write and publish some damn code, so that means a lot. Everyone else is either mining or marketing. And then there's someone who's either got well over a gigahash, or is writing and deploying *unpublished* exploit code to do this: 9Uueyr4T8dmzeZHNhYPjYdZbPpxmE9tUMh

If I wanted to line my pockets, I'd write my own automated counter-exploit code, and orphan those blocks, and not tell anyone about it. But it would probably cost me more to rent the hashpower than if I just bought a lot of catcoin.

What I would rather do is publicly explain to anyone who cares to listen how to modify a stratum proxy to get the network to be more stable *without changing catcoind*. But that will require coordination and some hashpower, and willingness to consider that maybe we should have a different way of looking at timestamps than Bitcoin did.

Now, I understand you and I disagree about the value of timestamps. So please, test your theories. Remove the 30 second minimum, and test it, live, and see what happens.

It irritates the hell out of me when everyone screams about "we don't have time to test this cause life", and then gets pissed off because I go and test something.

Or jeezus christ, go look at the travis-ci integration that mainline Bitcoin has for automatically building and testing every commit to the repository.

-- edit: addendum --
When I ran some code that nobody liked I thought Catcoin was an orphan, and needed someone to take it over. The fact that you all spent actual cash convinced me there was a community interested in doing something, and well, you know, being a community. The problem is that since then, nothing has happened besides a lot of infighting and FUDing.
And for the love of ceiling cat, post more cat pictures

Good to see those logos I threw together finally get some use. Again thanks to Ambethia for the actual cat mascot itself.

Just a few minor website updates to report.

The pools and exchanges listings at Catcoins.org are updated.

The new block explorer is linked as well.

We've pulled as many functions into catcoins.org that we can right now and will continue to strengthen that as the 'official site' it was intended to be.

Catcoinwallets.com will remain the primary wallet archive and will serve the block explorer, though we'll have redundant wallet storage at catcoins.org as well.  (And yes - the ugly page at catcoinwallets will be brought into at least the 20th century. )

The infrastructure is solid, documented, and has redundant back-ups for critical functions.  It's a good day.

Andy

Who's this 'we'? You certainly don't speak for me.

We've got a new pool in the family:

http://iegservers.com/

The partners behind this effort have been in the -dev channel and they're both motivated and responsive.

Welcome to the team, guys!

I'll add it to the Catcoins.org page next update.