• GreenIsMyPepper has quit
      • espes has quit
      • yorick has quit
      • CodeShark
        could we gain anything from rolling back only bad transactions and dependencies but keeping the block for PoW?
      • and for the other unaffected transaction chains
      • Starduster has quit
      • so we don't need to do full block reorgs
      • yorick joined the channel
      • dEBRUYNE has quit
      • a repudiation mechanism that punishes validators that didn't catch the error and rewards validators that find it :)
      • phantomcircuit
        CodeShark, speed of light issues iirc
      • CodeShark
        or provers that find it, rather
      • how is speed of light an issue? block propagation time is still far from relativistic timescale
      • the idea is if a particular block propagates, the network could tolerate the case of one or two specific transactions in the block being bad and not being immediately detected
      • the damage to the ledger as a whole would be minimal
      • granted the error is caught sufficiently quickly
      • perhaps it's a stupid idea...just pondering out loud :p
      • phantomcircuit
        CodeShark, how does having two blocks at the same height help you in anyway?
      • i dont get it
      • CodeShark
        no, I think we're talking about different things perhaps
      • you still have a well-ordered block chain
      • presumably with very short side branches
      • the motivation is not requiring full validation to still contribue PoW...but at the cost of risk...or something like that
      • that's to say, reorgs only affect a very small number of transaction chains
      • delegated validation...or blah, I don't know :p
      • nvm...perhaps it doesn't make any sense at all
      • yorick has quit
      • I guess at a more fundamental level, I'm talking about adding a little more fault tolerance to validation with economic levers that strongly incentivize cooperation
      • -- BotBot disconnected, possible missing messages --
      • -- BotBot disconnected, possible missing messages --
      • -- BotBot disconnected, possible missing messages --
      • [d__d] joined the channel
      • point is the more we can localize reorgs the less of the global history everyone needs to be aware of
      • or rather...what I mean is the the less that the network as a whole is impacted and disrupted
      • licnep joined the channel
      • wallet42 has quit
      • hashtag_ joined the channel
      • any chain invalidation is potentially costly...especially as the number of dependencies grows
      • which makes the structure EXTREMELY brittle
      • eventually we'll need consensus protocols that can be more fork-tolerant, I think ;)
      • chmod755 has quit
      • rusty joined the channel
      • I'm also speaking of transaction replacement in mempool and other such instances of invalidation chins
      • *chains
      • justanotherusr
        CodeShark: how do you localize a reorg safely?
      • CodeShark
        you can start by considering the diffs
      • what contracts remain in place?
      • rusty has quit
      • if only a relatively small portion of the network is affected by a change, perhaps it isn't necessary to make it matter at all to people who don't care about those differences
      • it's something like a revocation mechanism
      • but revocation or invalidation of a chain isn't necessarily free...it could come at a high cost
      • question really is: who's paying the bill?
      • and will there ultimately be convergence on consensus?
      • any such mechanism inherently carries risks...but if we can quantify the risks and allocate them algorithmically I think we're doing great :)
      • bendavenport has quit
      • rusty joined the channel
      • copumpkin joined the channel
      • justanotherusr
        how do you come to consensus on what should be in a reorg without a new consensus mechanism? :)
      • CodeShark
        you nuke it
      • basically, "kids, solve it yourselves or lose it all" :p
      • but everyone else's transactions continue to process on cue
      • if there's a fundamental consensus incompatibility, the system is screwed anyway...
      • but at least it could recover gracefully from relatively small glitches
      • it would be nice for it not to be quite so fragile - to be able to support a little more uncertainty
      • hashtagg_ joined the channel
      • hashtag_ has quit
      • Dr-G has quit
      • Dr-G2 joined the channel
      • and to find ways to contain disputes to be as local as possible
      • justanotherusr
        the solution is making a reorg unlikely by waiting
      • CodeShark
        from the perspective of an end user, sure
      • from the perspective of a protocol designer that's still a crapload of overhead
      • ruby32 joined the channel
      • justanotherusr
        Are you proposing there is another solution? I don't see one in the scrollback
      • CodeShark
        you can make it possible to invalidate specific transaction chains
      • Burrito has quit
      • ruby32 has quit
      • airbreather has quit
      • airbreather joined the channel
      • invalidate by signature, not by censorship :p
      • with a settlement
      • hashtag_ joined the channel
      • hashtagg_ has quit
      • belcher has quit
      • warptangent has quit
      • the real solution to RBF seems to be to provide for transaction repudiation
      • helo has quit
      • helo joined the channel
      • the existence of transaction X makes the existence of Y irrelevant
      • so if both get published on the block chain, X wins out...and Y potentially costs the miner money
      • warptangent joined the channel
      • this would allow for deterministic fee updates after-the-fact
      • and ultimately doesn't depend on any particular relay policy
      • or perhaps clever uses of locktime :p
      • nah, the real solution to fees starts by separating the fee as an explicit output :)
      • rusty has left the channel
      • moa
        it's explicitly the difference between the inputs and the outputs isn't it?
      • CodeShark
        yes...but the current structures are horrendous at making this easy to calculate :p
      • also, it's nice to have the local symmetry...the sum of all edges is zero
      • gielbier has quit
      • gielbier joined the channel
      • jaekwon_ has quit
      • ttttemp has quit
      • GGuyZ joined the channel
      • ttttemp joined the channel
      • jtimon has quit
      • currently, you really do need the whole mempool to make basic fee calculations :p
      • as well as the transactions that anchor it into the blockchain
      • it's stupid, really - lol
      • or perhaps stupid isn't the right adjective - let's say frustrating :)
      • it limits our tools to devices that have at least a certain amount of resources and a decent quality network connection
      • ThomasV joined the channel
      • to perform an operation that is practically trivial ;)
      • or, of course, it breaks our trust model (I sometimes wonder whether people in this space really care about this one anymore, though)
      • p15x joined the channel
      • Giszmo has quit
      • rusty joined the channel
      • GGuyZ_ joined the channel
      • GGuyZ has quit
      • GGuyZ_ is now known as GGuyZ
      • p15x_ joined the channel
      • p15x has quit
      • mjerr joined the channel
      • p15x_ has quit
      • ttttemp has quit
      • p15x joined the channel
      • ttttemp joined the channel
      • ttttemp has quit
      • ttttemp joined the channel
      • snthsnth joined the channel
      • TheSeven has quit
      • TheSeven joined the channel
      • GGuyZ_ joined the channel
      • GGuyZ has quit
      • GGuyZ_ is now known as GGuyZ
      • EasyAt joined the channel
      • frankenmint joined the channel
      • alawson joined the channel
      • www joined the channel
      • espes__ joined the channel
      • ThomasV has quit
      • mjerr has quit
      • espes__ has quit
      • rusty has quit
      • rusty joined the channel
      • ruby32 joined the channel
      • GGuyZ has quit
      • ruby32 has quit
      • espes__ joined the channel
      • hashtag_ joined the channel