Quote from: iCEBREAKER on July 01, 2015, 03:17:19 AM

Quote from: _mr_e on June 30, 2015, 03:18:05 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.

So they switch back to working on a full block if they aren't able to solve the block header in the time it takes to fully process everything?

Yes, that's right. Large miners always try for 0-tx freebie blocks while processing the pool, then switch to fee-grabbing mode when its done. The larger the block, the more (non-trivial) CPU time must be devoted to processing. It's a linear relationship, so 20MB blocks take 20 times longer. This is bad for an antifragile diverse/diffuse/defensible/resilient system designed to ~~remain functional~~ thrive under load, but not under bloat.

This delay is a form of propagation delay and thus drives up the orphan rate for miners with less resources. Afaik proportional increases in orphan rate are more costly than proportional decreases in hashrate, because the math is compounded (but diminishing) on each subsequent block of the orphaned chain. Thus this action doesn't appear to make economic sense unless it is explained as a lack of bandwidth and not a lack of desire to apply more of their resources to processing the txns than to hashrate. If it is bandwidth that is culprit, then it argues against larger block sizes.