Topic: Understanding why the call to rollback to v0.7 was made. - page 2. (Read 3298 times)

There shouldn't be such transactions, there can be fault at merchant end accepting double-spend transaction during fork time.

That's very bad.

https://bitcointalksearch.org/topic/a-successful-double-spend-us10000-against-okpay-this-morning-152348

Thank you for the post.Very insightful!

Good post.

Coinbase transactions (block rewards) from the roughly 25 orphaned blocks in the v0.8 chain.  It was important to resolve this before either chain was more than 100 blocks past the fork as then those coins would be spendable which would have been "chaotic".

Is there any transaction from abandoned 0.8 chain that was not included into legit 0.7 chain?

There seems to be some confusion on why a call to rollback to v0.7 was made so hopefully this will dispel some myth.

What happened?
The network hard forked on block 225433.  The block produced by a v0.8 node was rejected by at least some v0.7 nodes.  Contrary to initial claims this was more complex than just "the block was too big".  Blocks right up to the 1MB limit have been created and relayed without issue on testnet.  The BDB used by v0.7 failed to validate the block 225433 due to an inability to handle the number of tx inputs (not to be confused with block size or number of transactions) and rejected it.  This was an undocumented issue with v0.7 ("bug").

At this point a hard fork existed.  The mining nodes running v0.8 built upon the block 225433 and extended their chain.  The nodes running v0.7 rejected the block 225433 and eventually produced an alternative and incompatible block "225433a".   Since the deviation was uni-direction in theory v0.7 could eventually surpass v0.8 chain and re-unify the network however v0.8 chain had a significant majority of hashing power and that never happened.  By the time the issue was being analyzed v0.8 was over 5 blocks ahead and pulling further ahead with time.

Why was the fork time critical?
The users on v0.7 fork would never rejoin the network.  If all those users shutdown their clients and didn't accept any transactions there was no real risk for them.  However in time as hashing power fled the v0.7 fork remaining nodes would become more and more vulnerable to a variety of attacks.  Not just from a potential 51% attack, but also from accepting generated coins which were valid on the v0.8 blocks and even non hashpower related double spends. 

Why not just let "most hashing power wins"?
The "most hashing power wins" concept is used to solve splits/reorgs not hard forks.  It would create a bad precedent.  Essentially miners on v0.8 node would be leaving potential v0.7 users vulnerable to attack to save a few block rewards.  Since non-mining nodes (users) running v0.7 would never accept the #225433 block produced by v0.8 they would have no method to rejoin the longest chain short of upgrading.  v0.8 users (non miners) would accept blocks produced by v0.7.  By having a few miners downgrade to v0.7 it "reunified" the network in the shortest period of time with the smallest number of changes required.

What if developers/pool operators couldn't reach a consensus?  Would this have destroyed Bitcoin?
No.  The v0.8 chain would continue.  v0.7 users would be urged to immediately stop all transactions and upgrade.  Users on v0.7 clients would remain (increasingly) vulnerable to a variety of attacks if actively engaging in transactions until they upgraded.  Eventually all users would upgrade and the network would have continued with the v0.7 branch being an oprhaned sidechain.  The potential threat to active v0.7 users would make this less than optimal but the network as a whole was never in any danger of failure.

Was there a real risk of 51% attack on the major chain?
Not really.  The v0.8 chain had roughly 60% of the hashing power so to double spend that chain would require more hashing power than 60% of the Bitcoin network.  While 60% is less than 100% it is still a staggering amount of hashing power.  Any malicous actor building out a network that large wouldn't stop and wait for the remote chance that 60% would be enough.  Anyone with that amount of time and resources would be planning to outbuild the full network and attack.

What risks were there for users on v0.7 fork?
The largest risk would be from a 51% attack.  Initially the v0.7 fork was relatively well protected but had it been abandoned in favor of extending the v0.8 chain that hash power would have dropped over time.  The longer a user remained active on the v0.7 network the more danger they would be in.  Eventually hashing power would fall to a level where even a small pool could have easily executed double spends.  

Users on the v0.7 would eventually (100 blocks after the fork) be at risk of bring double spent by newly generated coins.  The coins would confirm but once they upgraded to v0.8 that transaction would never exist (as the block producing it also never existed).  v0.7 users who heeded the alert key warning and stopped all transactions were never in any significant danger.  

Lastly a smart attacker with good timing could have executed a "no hashing power" double spend against vulnerable v0.7 users (who continued to engage in active transactions despite the alert key warning).  A simplified version of this attack would work like this.  As hashing power on v0.7 fork fell, the block generation rate would slow and the number of transactions in the memory pool would increase.  An attacker could exploit this differential by creating a large number of low/ no fee transactions and wait for some of them to be accepted into v0.8 blocks.  The Bitcoin network will eventually forget about unconfirmed transactions as they get old enough to fall out of the memory pool.  Normally a client will rebroadcast an unconfirmed transaction periodically but in this case the attacker's client would see the transaction as confirmed.  The attacker could speed up this process but producing a large number of unrelated transactions.  When the tx in v0.8 blocks were dropped from the v0.7 memory pool the attack could double spend victims on the v0.7 network.