Quote from: cypherdoc on June 30, 2015, 07:49:09 PM

Quote from: Erdogan on June 30, 2015, 07:33:03 PM

Quote from: cypherdoc on June 30, 2015, 07:03:47 PM

Quote from: Erdogan on June 30, 2015, 06:55:44 PM

Quote from: cypherdoc on June 30, 2015, 03:06:19 PM

Quote from: _mr_e on June 30, 2015, 02:48:35 PM

Quote from: cypherdoc on June 29, 2015, 07:13:02 PM

blocks full again with TPS unacceptably high --> unconf tx set ~ 3000, more than double normal. pools taking defensive action. when will Blockstream devs do something?:

Why do you call this "defensive" action?

when Chinese pools face a series of what they consider to be "large blocks", which in this case means full blocks as you can see from the data, they automatically switch to mining "header" blocks with 0 tx's during the time it takes to process all the tx's in the preceding large block. this is b/c large blocks take a bit more time to process and check all the signatures so the argument goes that they can't afford to waste that precious time so for defensive purposed they just go ahead and start hashing the next block with only the "header" that in this case contain no tx's that might have included an invalid input from the preceding large block.

How about: When they see a new block, they immediately start mining based on that, with an empty transaction list. When they have removed the transactions in the last block from their own list of transactions to include, they mine with the new list. If a block is found quickly, it will be the empty block, else it will be a block with the new list.

You can see that the zero transaction blocks always come a short time after the previous. But they should not need minutes to establish a new list, so I am not sure.

Maybe some miner can chime in and say what really is going on.

your first paragraph was correct.

i'm not sure why the 0 tx blocks come so fast after the previous block. i keep asking this.

it's the sigop checking of a large block, i believe, that takes the extra time.

Well my first paragraph suggested a reason. They need time to construct a new transaction list. ~~When that is done, new transactions can more easily be added. The first list is the time consuming part.~~

Edit: The block finding is completely random. It is just as probable to find a block the first second, as after several minutes.

i know the probability is random. so why has every_single_one i've seen always been within a minute or two?

Because they need to frenetically keep mining to cover investments and cost. They have to start mining before they have a list of transactions, and sometimes they find a block in that time interval. A miner could easily confirm or refute this, if one is in the room.