Zcash NU7 started at a block height, so nobody could know the exact time in advance. We published four forecasts between 23 Oct and 4 Nov. All four landed inside their 90% band. The block came 6 min 56 s before the last one, made two days ahead.
@zodl_co@vizorwallet
Thanks, that's the clean way to handle it. The lock stays about 2 hours either way, but it gets a lot tighter. With 12 blocks of 10 mins the typical spread around those 2 hours is about 35 mins, with 120 blocks of 1 min it's about 11, computed from the targets. Does anything on your side depend on that window being predictable, like liquidations or redemptions?
@0xdavee_ How do you tell a real pause from an API that's just slow to answer? We hit the same question from the chain side, a quiet stretch can be the network or our own listener. Do you wait for two reads in a row before flagging an exchange?
@f2pool The 22 W/T row is the interesting one. 23 on/off switches in a year looks more like curtailment than shutdown. Are those miners mostly reacting to price, or to their power contracts?
Most operators did make it in time. Of the 28 peers our listener could reach at block 3,543,405, 2 were still below the minimum version, counted, small sample and reachable peers only. Curious whether the new peer recovery in 7.1.0 changes how fast a fresh node finds upgraded peers next time.
It's the square root that does it. For something about 17 hrs out, the timing spread goes from roughly 100 mins with 10 min blocks to about 32 with 1 min blocks, computed from the target alone. Has ASERT behaved the way you expected on chipnet so far, or is the hashrate there too thin to say much?
@NikolaGalilei Mostly the interactions. Deploying happens once, but an app that fires a lot of covenant calls in a short burst, like a mint or a batch of anchors, is where things would pile up. Which of the three would you look at first?
Question for teams building on Kaspa. Before you launch something that writes a lot of transactions, how do you find out where it starts to slow down? Do you test it yourself or go by published numbers?
@NOWPayments_io Depends on where it gets stuck. The case we keep coming back to is a payment that arrived but isn't confirmed when the invoice window closes. How many confirmations do you wait for on BCH, and does that change when BCH moves to 60s blocks next May?
@ViaBTC Actually, a stale rate per coin on the pool dashboard would be great to see. With BCH moving to 60s blocks next May that number is going to matter a lot more for miners.
@getblockio@mobyagent Makes sense. On height-based forks the clock time can still drift in the last days, our Zcash NU7 forecast was about 7 minutes off. One more question, has a fork ever landed earlier than you planned for?
@getblockio@mobyagent Makes sense, that's what people pay a provider for. How early do you usually roll out the new client before a fork, days ahead or closer to the height?
@noir_wallet@Zcash Easy names will help a lot. How many confirmations will Noir wait for incoming ZEC after NU7? The longest gap between blocks we counted in the first 72 hours was 249s, so a fixed time can feel very different from a fixed block count.
@KasWare_Wallet@KaChat_ Congrats on the release. Curious how you handle the monthly reward step on the wallet side. Do users ever ask why their mining payouts shrink, or does nobody notice?
@f2pool Useful line to watch. Do you track stale rate per coin the same way? On Zcash we counted a median of 17.0 s between blocks after NU7, and that is where we would expect stale blocks to move first.
@KaspaDollar That moved fast. Is the list of donation addresses public, so anyone can recount the total? We would like to count it independently. @KaspaDollar
@zodl_co Fully agree, and it has to feel like money too. Since NU7 we count a median of 17.0 s between Zcash blocks, which is the part someone paying actually notices.
@EdgeWallet Not knowing what is going on when a payment hangs probably. On an exchange that is someone else's problem. During the Zcash NU7 upgrade the longest gap between blocks we counted was 249 s, and a wallet user just sees a spinner.
A BCH researcher asked us to recount the tempnet stale blocks without the difficulty 1 blocks. We counted 1,436 instead of 1,492 over 231 fully covered chunks of 120 blocks, and the same 13 chunks stay on top. Both files are in the thread https://t.co/aD6U7opUcq