Proposed change for the 0.8.6 release: increase the default block size by to 750K, and free transaction area to 50K.
There is a more conservative change in 0.8.6:
See https://gist.github.com/gavinandresen/7670433 for current thinking on how fee-handling will evolve over the next couple of releases.
34@@ -35,10 +35,14 @@
35
36 /** The maximum allowed size for a serialized block, in bytes (network rule) */
37 static const unsigned int MAX_BLOCK_SIZE = 1000000;
38-/** The maximum size for mined blocks */
39+/** Obsolete: maximum size for mined blocks */
40 static const unsigned int MAX_BLOCK_SIZE_GEN = MAX_BLOCK_SIZE/2;
Your description says increase by 100kb, but as far as I can tell it’s only a 50kb increase? Default was 250kb after this change it’s 300kb. Which was the intention?
Luke-Jr runs with 1mb and BTC-Guild runs with 500 iiuc so it seems a 50kb bump may be too conservative given that we’re now routinely over 1 tx/sec:
https://blockchain.info/charts/n-transactions
At 100,000 txns/day (let’s be optimistic and hope that the new users stick around) that’s an average of ~695 transactions per block, at about 200 bytes average that’s 139kb. (does anyone have a graph of avg tx size?). So 300kb means the average block is about half that. But variance means blocks don’t come every 10 minutes, it’s pretty common to have to wait 20 or 30 minutes. Any time that happens the new soft limit will still be hit and the queue will start to back up/mempool will start to expand.
It’d be nice to have a rule of thumb for block size vs average tx rate, but I guess if we say going an hour with no block is pretty common, but 2 hours isn’t, then soft block size should be 6x daily rate/144 so that when the network waits an hour to get a block, the backlog can still be cleared. That would imply 140*6=840kb at current traffic levels.
