Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

OptiFi Program Incident Report

medium.com

Post reply on HN
Print view View original

141–150 of 158 posts

Re: OptiFi Program Incident Report

#141

Post by korginator »

The report reads like the OptiFi folks were trying to figure out how the different bits of the puzzle on the fly. Did these folks even know how to use their tools?

Towards later in the article they say "Here it turned out that we didn’t really understand the impact and risk of this closing program command line. ‘solana program close’ is actually for closing the program permanently and sending the SOL tokens in the buffer account used by the program back to the recipient wallet."

What?!

Re: OptiFi Program Incident Report

#142

Post by pjc50 »

benreesman wrote: post #75
tinco wrote: post #66

Earlier quoted context omitted.

Was it Cambridge Analytica that was going to pay 50MM for that data? ;)

No, it was features for a P(click|impression) model, which are surprisingly mundane: it is basically impossible to care much either way, let alone be outraged, if you look at the actual features used in these kinds of things and it’s not explicitly your field. It’s not a microphone listening on your smartphone! You can even read about it in this paper which publicly described the system contemporary to that event: ht…

I'll just file it under "best minds of our generation are working on making people click 3% more often on ads".

Re: OptiFi Program Incident Report

#143

Post by pjc50 »

Vespasian wrote: post #130
Yizahi wrote: post #106

Earlier quoted context omitted.

First to market is only part of the problem. Another self inflicted problem is that execution of those "smart" "contracts" cost tokens (so essentially money). This incentivizes developers to write the shortest possible code, without any "fluff" like tests or additional checks or more verbose style. And the immutability problem also increases severity of the problem - you either deploy immutable "contract" signifying…

Immutable code is a poor replacement for trust in a company. It shifts the trust from the company as it is now to the same company (and their capabilities) sometimes in the past. Preferably the company is operated completely anonymously. It seems like the crypto community would do anything to avoid the legal/regulatory system and it's established processes of operation requirements, insurance and liabilities. Doing t…

The closest I can think of is the TSB migration failure: https://www.theguardian.com/business/2019/nov/19/tsb-it-melt...

That also involved failing to have a viable rollback plan. It affected so many people so badly that questions were asked in Parliament; TSB were down for a week.

Now with cryptocurrency we've disintermediated the bank to produce a much better solution: your contract can be down forever and there's no Parliament to ask questions in.

Re: OptiFi Program Incident Report

#144

Post by eloff »

I've watched crypto for years with interest, I've worked in the field, speculated in it with my savings, used it for international transfers, spent it, etc. That is to say I'm not some armchair critic.

I find myself lately viewing crypto as mostly a bad idea, or a bad implementation of an idea with potential, or even outright as a Ponzi scheme.

These smart contacts are really a bad idea. We have contracts in life with lawyers, these tend to work out pretty well because we can use our human intelligence to arbitrate them. The idea of code as law is really dumb. I say that as a programmer who works on code all day. Computers are dumb like your calculator. They do exactly what you tell them to. Which sounds great until you realize how hard it is to describe what you want in a way the computer can understand. That's why I make a good salary and people think my job is difficult. The reality is computers often do what you tell them, but it's not what you wanted. Rather you have the instructions wrong or you made a mistake. That's why software is riddled with bugs. Avoiding that is very, very, very difficult. Nobody gets it right. These crypto finance bros will definately not get it right. And when they get it wrong, there's no human, no court to appeal to. How's that an improvement?

Re: OptiFi Program Incident Report

#145

Post by pjc50 »

jdasdf wrote: post #78
spaceman_2020 wrote: post #39

This is one of the biggest flaws in crypto. Small errors can erode hundreds of millions of value. That's when most crypto companies right now are very small. Can you imagine the chaos if crypto were to actually become big and dev count were to grow to 1k+ people spread across multiple offices? There's definite efficiency gains with crypto (a dex like Uniswap can do massive volume with very few developers for instance…

> This is one of the biggest flaws in crypto. Small errors can erode hundreds of millions of value. That's reality. Same thing would have happened if they had put all that cash on a boat, and accidentally sunk it.

This is a deeply stupid analogy, because as everyone is pointing out .. why would you do that? Why would you choose to build a system in which this is a risk, when it's not actually imposed by the laws of nature?

But there were a few instances in WW2 where large amounts of value had to be transferred by boat, and in at least one case sunk by enemy action. HMS Edinburgh: https://www.warhistoryonline.com/instant-articles/hms-edinbu...

Re: OptiFi Program Incident Report

#146

Post by klabb3 »

throw101010 wrote: post #110
birracerveza wrote: post #101

Earlier quoted context omitted.

Multiple DeFi platforms exist. Some are mostly decentralized, others only use DeFi as a marketing buzzword. Some of these are actually centralized, but it doesn't mean all of them are.

In agreement with GP, Solana and other very centralized networks should have to earn their way into the term DeFi. It's hard to quantify but a network that can be "paused" by few devs and relying on the existence of a registered company clearly don't fit this term.

Yeah. I don't get why bother at all with the blockchain when a simple dB would be even superior, then they can prototype away and migrate to decentralized later, if anyone even wants that. Decentralization shares the "weakest link" property with cyber security: you don't get the benefits of you only do half of it.

Re: OptiFi Program Incident Report

#147

Post by scotty79 »

bhaak wrote: post #5

For all its flaws (and there are countless) the one thing about the show cryptocurrency space that stands out to me as a programmer is that programming errors can be suddenly very costly (granted, in this case it was more of a DevOps blunder). Being able to very easily put a price tag on sloppy programming is intriguing to me.

When I first found about cryptocurrencies it struck me as it was the first time in the history where data on my computer were directly worth something. Which suddenly made security worth something.

Re: OptiFi Program Incident Report

#148

Post by rjh29 »

selfhoster11 wrote: post #59
rjh29 wrote: post #40

Earlier quoted context omitted.

I mean that won't work, because --preserve-root is on by default. In the same way, eventually crypto may acquire the same checks and balances that other industries have, but it'll take a lot of costly mistakes to get to that point.

Preserve root exists now, but it didn’t exist in the past.

That's my point. This is the early days for crypto too. I'm not a fan of crypto, but it's clearly going to learn from these mistakes and develop the same kind of safety guarantees over time.

I mean look at online banking - in the UK it's only in the last few _years_ that we've been required to do 2fa when buying things online. The earliest online banks didn't even have 2fa to log in and move money around, and at some point HTTPS didn't even exist so everything was sent in plaintext. SMS is still mostly plaintext. Things evolve.

Re: OptiFi Program Incident Report

#149

Post by vineyardmike »

pjerem wrote: post #127
vineyardmike wrote: post #62

Earlier quoted context omitted.

> I and OP are saying, the future will look more like crypto looks today. Not a bright future. I don’t agree. I trust that loopholes like that will be slowly rectified with legislation if not present today. In fact, it’s likely that crypto will (problematically?) be heading the same way. There was a recent case of a crypto buff who found a bug in some project and made off with a few $Million and I think the courts sa…

> he could be arrested and expected to return it, just as if he made off with cash. You have cash in your house and someone breaks in to steal everything. The insurance will _maybe_ cover your loss given that you secured it with basic security. _Maybe_ the police will investigate and arrest the burglar. But even in this situation, you were better off putting your savings in any bank account where any fraudulent trans…

By cash I meant fiat. But yea crypto adds a lot of risk to storing your monies.

You can insure the crypto you have. It’s probs expensive to the point of being not worth it.

People investigate crypto hacks though. And if the perpetrators are in a jurisdiction that you have some legal availability to you can totally use legal means. Basically any western nation will allow such a suit.

https://www.coindesk.com/policy/2021/12/22/teenage-suspect-i...

Re: OptiFi Program Incident Report

#150

Post by benreesman »

pjc50 wrote: post #142
benreesman wrote: post #75

Earlier quoted context omitted.

No, it was features for a P(click|impression) model, which are surprisingly mundane: it is basically impossible to care much either way, let alone be outraged, if you look at the actual features used in these kinds of things and it’s not explicitly your field. It’s not a microphone listening on your smartphone! You can even read about it in this paper which publicly described the system contemporary to that event: ht…

I'll just file it under "best minds of our generation are working on making people click 3% more often on ads".

Bernoulli distributions are everywhere, and as a result binary classifiers are everywhere. COVID tests (any binary medical diagnostic), fraud/no-fraud credit-card processing decisions, loan decisions, spam filtering, “toxic” tweet management and other sentiment analysis, it just goes on forever.

All of these things benefit from advances in binary classifiers, and have therefore been massively subsidized by click prediction R&D. Maybe those benefits outweighs the costs, maybe they don’t.

We have GPUs because shoot-em-up games were “the big thing” in the 90s. Net win? Personal preference. But entertainment if one form or another often pushes advances in computing. The BOPR stuff from the linked paper originated on XBox Live matchmaking.

It’s perfectly fair to be like: I don’t like the ad-funded internet. There are tons of plausible reasons why someone wouldn’t like the ad-funded Internet, but most reasonable objections to it are of the form: “my internet experience is worse than it would be without ad funding”, past that you rapidly wander into “other people’s preferences are inferior to mine”.

Post reply on HN