Parity's Wallet Bug Is Not Alone
hackingdistributed.com
hackingdistributed.com
Blockchain contracts for real money need at least the assurance level of database transactions. The DAO attack involved causing a transaction to happen only in part; one part happened, then the transaction aborted. That needs to be prohibited by the underlying system. As with database transactions, either everything goes right and the transaction commits, or something goes wrong and the data is unchanged.
But forgetting that the wheel is not only not reinvented, but the market or industry or hype chain has not yet advanced to appreciation of wheels, is potentially a costly error.
This is how I explained to myself why I found it so hard to first get a understanding of what it was about : nobody wanted to say what it is.
Many people seem to get irrational if they fear to lose their virtual money.
For that matter, I always look closely how people act when the price dips due to some insignificant drama. You can assess beforehand if you want to throw money towards a project depending on those people.
Because when the next real crisis hits, and believe me, in cryptocurrencies there's always a next one, those people won't act better in a real crisis.
And then BitGo goes on to be partially responsible for a $320,000,000 loss (current value) that almost destroyed BitFinex.
It's just sad that so much fighting and ego has prevented technical collaboration. I'm a supporter of the Core devs but Emin is a genius who should be respected.
Emin has a bad track record of lying about his work.
For example, he claimed in the announcement(1) for his Teechan payment channel scheme that it could do 2480 transactions per second, but neglected to mention that it achieved that by failing to actually write those transactions to disk and storing them in RAM only. If your computer crashes, you can lose money in Teechan. This is not unlike advertising the high performance figures achieved by making a new RAM-only database, without advertising the fact that you achieved them by storing everything in RAM.
Similarly, Emin's announcement and paper also gave the impression that payment channels weren't currently possible to implement on Bitcoin without segwit, when in payment channels with similar properties to Teechan have been possible to implement for multiple years now via BIP65. Oddly, BIP65 is cited in the Teechan paper, but for an unrelated reason.
Then there's how Emin's PR around that announcement presented Teechan as something that could be implemented right away as a replacement for segwit via Intel SGX, without mentioning that Intel required SGX users to get licenses to use it in production, and Teechan didn't have one.
I could go on, but that's just a single project... The sad thing is Emin is often right and can do good work, but with a track record like that it's no surprise that people aren't interested in working with him.
1) https://web.archive.org/web/20161223143716/http://hackingdis...
>For example, he claimed in the announcement(1) for his Teechan payment channel scheme that it could do 2480 transactions per second, but neglected to mention that it achieved that by failing to actually write those transactions to disk and storing them in RAM only.
First of all, our actual deployments showed that Teechan can do more than 30,000 tx/sec across the Atlantic, in fully fault tolerant mode.
Second, the Teechan paper made clear exactly how we reported the initial, unoptimized numbers, which is precisely why you're here writing these bogus critiques.
>Emin's announcement and paper also gave the impression that payment channels weren't currently possible to implement on Bitcoin without segwit
We learned after publication that there are rumors of a "Lightning Network" implementation without Segwit. Unless you can point to a protocol specification, however, these remain unbacked and uncharacterized assertions, of the kind "I believe I can fly."
>Then there's how Emin's PR around that announcement presented Teechan as something that could be implemented right away as a replacement for segwit via Intel SGX, without mentioning that Intel required SGX users to get licenses to use it in production, and Teechan didn't have one.
This is HN, and I usually try to be more diplomatic, but this is stupid. First, Teechan is a protocol, it isn't tied to SGX. We are working with HSM vendors to deploy it on non-Intel hardware. Second, Teechan does NOT require the users to obtain licenses, it just needs to be signed by an entity. Third, Peter Todd has no idea about the nature and scope of academic work and how it differs from industrial deployments, which explains why he feels so threatened as to indulge in personal smears.
I could go on, but tearing down a known Bitcoin troll is pointless.
I'd appreciate if you linked to those actual deployments, something with a description of how they worked.
As you know, actual remote attestation capable hardware has severe limitations on persistent anti-replay mechanisms due to the fact that they are implemented in hardware. Your Teechan performance figures reported in your paper and initial announcement were recorded with those counters implemented in RAM, in such a way that a power outage would put users in a position where they can lose funds. Meanwhile, the actual SGX anti-replay counters that are currently available limit counter updates to as infrequent as multiple minutes between updates.
> We learned after publication that there are rumors of a "Lightning Network" implementation without Segwit. Unless you can point to a protocol specification, however, these remain unbacked and uncharacterized assertions, of the kind "I believe I can fly."
Payment channels != Lightning network.
Specifically, you presented Spilman payment channels as state of the art, when you are well aware of the fact that they have been made obsolete by CheckLockTimeVerify/CheckSequenceVerify payment channels.
Teechan as presented in your paper that I was criticizing wasn't a Lightning network competitor, it's a payment channel competitor. So obviously an apples-to-apples comparison would be appropriate.
> Second, Teechan does NOT require the users to obtain licenses, it just needs to be signed by an entity.
What do you mean by "signed by an entity" in your statement here? To be clear, by "SGX users" I am referring to developers of SGX-using products, not end-users.
Also to be clear, you claimed - based on the fact that SGX was widely deployed - that "[Teechan] side-steps a controversial proposal to change the underlying Bitcoin protocol, and provides all of the much-touted benefits of Lightning Networks today, without having to modify the base protocol at all." (emphasis mine)
To say that without making it clear that SGX is not in fact something that can be deployed on a whim due to the necessity of getting a license to use a SGX application in production is quite dishonest, especially in the context of the Bitcoin segwit debate that you positioned Teechan within.
> I'd appreciate if you linked to those actual deployments, something with a description of how they worked.
I'm not sure if this is it, but it looks like something was recently put online (seems to match the authors names): https://arxiv.org/abs/1707.05454
> Your Teechan performance figures reported in your paper and initial announcement were recorded with those counters implemented in RAM, in such a way that a power outage would put users in a position where they can lose funds.
I don't think they disputed that, in fact iirc, they made that quite clear and even under failure cases where a user's CPU loses power, funds are not completely lost; two options remain, either the counterparty can settle the channel (meaning no funds are lost), or the refund tx comes into play. Of course this isn't ideal, but the point remains, for channels where the difference between the most recent state and the refund state is not enormous, this deployment model can make sense. If you want to avoid this, it can be circumvented with backup power generators etc. But they never tried to hide this fact. If you read the paper, they were clear that Teechan did not use hardware counters.
> Teechan as presented in your paper that I was criticizing wasn't a Lightning network competitor, it's a payment channel competitor. So obviously an apples-to-apples comparison would be appropriate.
Surely for an "apples-to-apples" comparison as you request, CLTV/CSV payment channels don't even compare with LN/Teechan: CTLV/CSV are only unidirectional and they suffer the exhaustion problem. In addition they can only be funded by a single party, so to compare them to bidirectional payment channels like these would be unfair too, no? Also, if you want to compare LN/Teechan pre-segwit, you have the same problem: LN payment channels can only be funded by a single party, in addition, they also require monitoring the blockchain (at the moment). I'm not saying Teechan is the better, I'm saying it has it's own set of advantages/disadvantages and as someone who seems to be pushing back against Teechan a lot, you should know these trade-offs?
> To say that without making it clear that SGX is not in fact something that can be deployed on a whim due to the necessity of getting a license to use a SGX application in production is quite dishonest, especially in the context of the Bitcoin segwit debate that you positioned Teechan within.
Granted, although this might be the biggest challenge in deploying Teechan today, it's by no means impossible: I know of a few companies who have licenses with Intel. Intel didn't create this stuff to sit on a shelf. And even if you ignore Intel, it seems to me that several companies are now looking into deploying payment channels with trusted hardware. e.g. I'm sure you are aware of Blockstream :)
To summarize, Peter, sometimes you have to give credit where credit is due. I know you might have personal differences with Gun, and I don't want to get between you, but for outsiders who don't really have much technical insight here, you seem to be twisting the facts a little? Sure, you might not agree with the work, but hey, the company you work for seems to like it enough to pay someone to do it (http://jobs.khoslaventures.com/jobdetail.php?jobid=698427)
I recently wrote a tool to help find bugs like this:
https://ericrafaloff.com/introducing-the-solidity-function-p...
He probably just wanted to share his holiday photos.