Quote from: radus on October 03, 2014, 08:29:08 AM

Quote from: aptcoin on October 03, 2014, 04:15:30 AM

Quote from: radus on October 01, 2014, 08:25:52 AM

Quote from: woot on September 30, 2014, 07:09:26 PM

Quote from: radus on September 29, 2014, 05:15:44 PM

Sorry for the double-post in advance.

Some of you might have noticed the wallets inability to sync for the last 6 hours or so - that Is entirely my doing.
No, the wallet is not broken and it is still perfectly capable of synchronizing - there just hasn't been generated any blocks for the last 6 hours.

I went through the code and noticed that the new algorithm for difficulty retargeting is, although whole-heartedly meant to be better... but in fact it's not.
Doing a few preliminary tests (offline) showed that difficulty retargeting is extremely agressive when a lot of hashing power is thrown at it. I wanted to bring this into light while also showing the downside to this new algorithm.
So I threw all my hashing-power against the network and thus got the difficulty for a n-factor 13 block up to something in the vicinity of 1.7k compared to the roughly 220 it was before. Don't worry, when the current block is solved, difficulty should be down quite drastically - although that might take some time.

TL;DR - devs... fix your new diff. algo.

It looks like everything is back to normal now, But good to know there are people testing this stuff out! So do you think this is worse than before the retargeting or not?

Definitely worse, but it can become better with a little tweaking.

Radus, thanks for testing :-)

I'm curious why you think this algorithm is worse than before though. For example, in the past we were stuck on blocks (i.e. a handful) for many days, and that has not happened again and our network continues (or would you disagree?).

Glad that you noticed the algorithm is aggressive, but from what you describe, it sounds like you're trying to make the algorithm re-target up, and you should see that we specifically dampened the re-target up. Where we're really aggressive is targeting difficulty back down, which is why the network is chugging along despite this test. We're not sure that we fully understand the scenario that you're testing, but we did test ramping up hash power and seeing what happens after a stuck block quite a bit. Can you help us to understand better what you did and why it caused problems? From here things appear to be the expected outcome, but it sounds like we missed something. What would you suggest?

I'm gonna have to say sorry here. The preliminary test was over the course of a few hours. I've done some testing now over the last 2 days and I've gotta say, it's definitely better, but difficulty doesn't scale quite as well as I had hoped (for each n-factor). It's still easy for me to ramp up the difficulty and let it hang at a high n-factor for a few hours, but then it drops significantly more than I previously believed. I don't know what went wrong in my previous test. Disregard my previous statement. The new algo is working much better, but I still feel it could handle a little tweaking - but that can wait for another required hardfork. Well done devs

Ok, thanks for clarifying (and again, thanks for testing this). I would argue that ramping up difficulty and having a stuck block can happen for any (bitcoin derived) cryptocurrency (although it's easiest to execute on lower hashrate networks). We're all for tweaking though, specific to how our system works, and wanted to see how this does over time. As you alluded to, keeping the hard forks down to a minimum is something we're striving for, but we are open to debating if one is needed in the future.