> Mr. Marino did not challenge the underlying facts of the case. Instead, he said they added up not to a crime but to a simple breach of Goldman’s confidentiality policy.
> “There’s no doubt Mr. Aleynikov did something wrong,” he told the jury, stopping occasionally to sip from a water bottle he clutched to throughout the closing statement. “But this is a very badly misguided case.”
There has been a lot of misinformation surrounding this case. I think its well established now that Sergey Aleynikov did indeed take more source code than just his open source contributions with him, his lawyer admits as much.
So I guess the take away is on your last day of work, don't just zip up and ftp any code from your company, it's just not worth it.
Actually I think the real takeaway is that if you are really good at your job, your company will pay more attention to you, especially when you leave.
Sergey was a big part of the HFT team at Goldman, he literally rebuilt their infrastructure and his leaving was a huge deal to the team, and he wasn't just leaving to go to say Google, he was going to a new HFT outfit that would compete with Goldman.
His actions would be like Mark Lucovsky leaving Microsoft with some windows NT kernel code he wrote, or Jeff Dean leaving Google with parts of the deep learning code to go to competitors to work on the same thing. I'd be willing to bet that both google and Microsoft would get lawyers involved in both those cases.
If you want to make open source contributions while at work join a company like, um...., Microsoft?
EDIT toned down the language calling him the entire HFT talent at Goldman, it was a bit over the top:)
I was actually shocked to find out that there are actually a few programmers that does this. I don't know what is the reason but it is tantamount to stealing.
Thank you for pointing this out. I wasn't aware that his original narrative was inaccurate, him having taken proprietary source code that he developed makes this outcome substantially more tolerable.
The problem here isn't so much with the facts, it's with the prosecutors and polices eagerness in pursuing it.
They only mention it in passing, but the reason he got all that evidence thrown out was essentially that the police were acting on very little else than Goldmans assertion.
Then he got acquitted on what was if you want to a "legal loophole", but still (and even that only happened a few steps up on the jurisdictional ladder, even though the law clearly didn't apply to him).
And then they charged him again, with essentially the same crime. So the same people that can't manage to put a single Goldman dude behind bars are all too happy to destroy lives on Goldmans words.
I'm not sure I follow. Your comment starts by saying the problem isn't with the facts --- tacitly agreeing with 'chollida1 that wrongdoing was established --- but ends with the assertion that Aleynikov was convicted "on Goldman's words". Can you help me understand how both of those things can be true?
Not the OP, but I read his comment as frustration with "One law for rich, another for the poor". One can agree that Aleynikov should be convicted, but hope that the law should pursue people who play fast and lose with public money with the same zeal they shown here.
Sure. But perhaps some of that frustration should be directed at legislatures that are too afraid of disrupting the economy and being blamed for job loss to criminalize the shady stuff giant banks do? Because one reason there's "one law" for people like Aleynikov and another for bank execs is that theft is black-letter illegal, and whatever-it-is we want to blame GS for might not be, yet.
One bias prosecutors definitely have is taking cases they think they can win.
> the police were acting on very little else than Goldmans assertion
That's a fair statement by the OP. The external criticism has always been there was a lack of a full investigation from a neutral position by the state, as investigators depended on Goldman Sachs people to determine the crime.
Outside neutral parties weren't consulted AFAIK. Which is typically the case when sophisticated crimes take place - such as bringing in private industry experts.
Whether or not that would have resulted in different charges or handling by the police is up for debate. But it is certain that the worst of the original charges were found to be baseless, prosecutors were over-aggressive, and most of the evidence was thrown out because they didn't have a proper basis to conduct search warrants.
Criminal proceedings are adversarial. The defense brings in experts specifically to rebut the charges. The prosecution brings in experts to support them. Where do "neutral experts" fit into this process?
Neutral in that the police and prosecutors verify that the claims by the victim are actually true otherwise anyone could use the police/prosecutors for destroy someone that they didn't like.
Much in the same way that ICE has acted like the enforcement arm of the RIAA/MPAA when taking down domains. Barely (or not at all) verifying that the domains were being used for infringement (instead just taking a list given to them by the RIAA/MPAA and acting on it blindly).
To the extent that's true, it's not really material to this case, right? His own lawyer stipulated to the facts. The dispute was a matter of law, not of fact.
I agree with the prevailing sentiment on HN that (a) it's too easy to get an indictment (the grand jury is supposed to be adversarial against the prosecutors and isn't) and (b) prosecutors bias --- my belief is: primarily bias towards cases they can win easily --- has subverted the goal of prosecutorial discretion.
But the facts of this case suggest a theft that could reasonably be valued at extremely high numbers. For it to be heard in court doesn't sound like a miscarriage of justice.
I agree. I'm just stating that I feel like people are unsettled by how it seemed like the victim supplied all of the evidence to the police/prosecutors and there was a general, "oh well, if you say so," feeling to the relationship between them.
Doesn't that describe basically every theft case ever? Someone breaks into my house and steals my TV, the same thing's going to happen. You don't think maybe people are just reacting knee-jerk to the name "Goldman Sachs"?
If someone breaks into my house, the police come into the house to investigate. I don't gather all of the evidence for them while preventing them from entering the premises. When has a theft occurred where the police just accepted all of the evidence from the victim, but didn't really do any evidence-gathering themselves?
It's one thing to commit a crime, and another to be put on trial* for it. The crime is his own, but I'm pretty sure the two trials are of Goldmans making. You don't get two trials and a law to boot from a single fervorous prosecutor.
They would have probably already dropped this case the moment most of the evidence was ruled out and called it a day.
Sorry but no matter how good he was, it's ridiculous to say he was the entire talent. I'm in a good position to know: I worked for several years within the group he was a part of (right after he left) and there were tons of terrific people there.
It's far more ridiculous to threaten and abuse our abject criminal system to reign in (that is if we assume the theft and not OSS happened) a talented employee who decided to jump the boat and join competition. Make no mistake, the people (wish their names were commonplace!) spearheading the legal pursuit on behalf of Goldman Sachs are evil - of the Hollywood/MPAA kind. Be wary of joining such Corporations, especially if you're a creative developer.
He wasn't convicted for violating a noncompete. He was convicted for taking Goldman property on his way out the door. His own lawyer stipulated to that in his summation.
I don't think it is well established that we he took was an algorithm or any such thing.
Michael Lewis wrote a big article in Vanity Fair which I suggest you look at. He gathered a number of experts who looked at the evidence presented at his federal trial and concluded that what he took was modified open source code to run the queues, do messaging, etc. Boiler plate code to create the infrastructure.
This is what GS is calling "proprietary source code", since he made the changes on GS dime therefore it is proprietary to them.
That's a common misunderstanding. According to the Second Circuit:
In addition to proprietary source code, Aleynikov also transferred some open source software licensed for use by the public that was mixed in with Goldman’s proprietary code. However, a substantially greater number of the uploaded files contained proprietary code than had open source software.
The Second Circuit is not an expert on software. It is an expert in law and modified open source code becomes proprietary code in eyes of the law for the purposes of GS employment agreement. They're not parsing how GPL or some other license might have applied.
I encourage you to look at Michael Lewis's article[1]. This passage is relevant: "As Goldman hadn’t permitted him to release his debugged or improved code back to the public—possibly in violation of the original free licenses, which often stated that improvements must be publicly shared—the only way to get his hands on these was to take the Goldman code. That he had taken, in the bargain, some code that wasn’t open source, which happened to be contained in the same files as the open-source code, surprised no one. Grabbing a bunch of files that contained both open-source and non-open-source code was an efficient, quick, and dirty way to collect the open-source code, even if the open-source code was the only part that interested him."
Once again: his own lawyer stipulated these facts in his criminal summation. It's not in dispute; it is undisputed. He did not simply taken open-source software. You can add this to the list of things Lewis got wrong; it's a large and growing one.
As Aleynikov concedes in his motion papers, the code he took from Goldman Sachs included a "purposefully designed" portion of the Goldman Sachs "proprietary, custom-built trading system." Indeed, the evidence showed that Aleynikov took a significant percentage of the proprietary source code for that system. While Aleynikov attempted to show that there was open source code embedded within the proprietary code and to identify the files in which that might be true, his expert witness was only able to identify one file among those taken by Aleynikov that both bore a Goldman Sachs copyright banner and appeared to contain open source code.
This does not contradict what Lewis writes or what I mentioned on my first post. The "proprietary, custom-built trading system." has a lot of infrastructure code to handle the incoming volume, ensure nothing gets dropped, and trades get evaluated. He doesn't seem to have taken the "decision making" part which would have been their secret sauce.
Sure, the modified open-source code might have been protected by GS employment agreement but that doesn't mean that it was their secret sauce. It was infrastructure/boiler plate code.
Additionally the fact that a code has a GS copyright banner doesn't mean that it was all brand-spanking-new GS-unique code. Even Lewis's article talks about the fact that a single file had open-source code and modified open-source code, aka proprietary code. This subtle language difference is exactly how a lawyer confuses the jury and it is difficult to counteract. Remember the SCO Unix v Novell fiasco!?
Lewis brought together experts who reviewed the evidence on record, i.e. the actual source codes submitted at evidence at his first trial and they reviewed it and found it to be derivative open-source code to run queues, and run infrastructure.
Have you read any of the court filings? Can you point to the one where it's established that Goldman's proprietary code is merely modified open source code? Even Aleynikov doesn't appear to be arguing that. Both sides in the trial had expert witnesses --- the Second Circuit didn't merely conclude on its own that the code was proprietary (in fact, they can't do that: appeals courts resolve disputes of law, not of fact).
Your source, by the way, does not support your assertion about Lewis contracting experts reviewing the source code presented as evidence during the trial. Which makes sense: what would make you think Lewis would have access to it in the first place?
Can you point to one instance where GS argues that it was the secret algorithm? They merely say code related to the trading platform. Lewis had access to the source code because it is presented as evidence in a trial and therefore matter of public record.
No one from GS has said that what he has taken was an algorithm or anything related to decision making. Not one person.
I never said the Second Circuit resolved a dispute about it being proprietary code. The defendant's own lawyer doesn't dispute that it was proprietary. But merely being proprietary is not sufficient for a trade secret violation. If you modify a config file for Apache that modified config file becomes proprietary but it is most likely not an essential trade secret that is the core of your business that someone else couldn't have come up with something similar to solve your problem.
GS lawyer's say that proprietary code WAS trade secret and essential, whereas defendant says no it was modified open source code used for queueing etc. The jury is not sophisticated enough to differentiate between the two. See any patent case where a BS software invention is held up as unique.
To address his lawyer's admission... as I said modified open-source code becomes proprietary so the lawyer isn't wrong. That is why he said that he violated the confidentiality agreement but that isn't the same as stealing trade secret. A company's trade secret isn't the same thing as proprietary code.
One of the legal qualification of trade secrets is that it is a method, process, etc. that isn't known or reasonably known by others for gaining advantage over competitors. It is their secret sauce.
I don't see anything here that suggests he took the ALGORITHM or DECISION MAKING process, or any such thing.
Can you point to the specific things you're relying on to determine what ALGORITHMS and DECISION MAKING processes may or may not have been taken? I'm not sure why I'm meant to be too concerned about which parts of the evidence you personally have or haven't seen.
Besides, you're reasoning about this on a faulty premise. This is a common misconception that people outside the industry have regarding HFT. Let me assure you, often the decision making is trivial, the answer to the question of what to do can be quite obvious. Then it's more about who can do it faster, and that's all about the plumbing. Which he stole.
Among the HN crowd I am sure you will find people on both sides of this issue: some people will think that GS is not working on 'good' problems that better the world, therefore, it is 'shady' to work there; and others will argue that GS is abstractly bettering the world through efficient resource/money allocation, especially for developing countries.
>others will argue that GS is abstractly bettering the world through efficient resource/money allocation, especially for developing countries
This would be true if GS didn't scam people, which they do. I accept the investment allocation argument, but I'm not convinced that it's valid for GS specifically.
EDIT: Holy shit, that controversies section just goes on and on. I just checked, and it's 7000 words long. This company's insane.
Anything that even smells like the financial and banking industry is considered shady by a lot of people, not just programmers, after our Great Recession. A necessary evil at best, and for programmers one that pays very well. In any case, it was a good zing.
I first heard about the programmer mentioned in this article in Michael Lewis' "Flash Boys," which goes into great detail about how places like Goldman let high-frequency trading firms into their "dark pool" to make money off arbitrage by being able to make orders faster than anyone else, and then steering their customers into their dark pools rather than executing the trades in the way most beneficial to their customers. I find that sort of thing to be rather unethical at best -- if I hire a broker, I expect them to buy and sell stocks in the way that best benefits me, not whichever one benefits some high-frequency trading firm the most.
> if I hire a broker, I expect them to buy and sell stocks in the way that best benefits me, not whichever one benefits some high-frequency trading firm the most.
This is not just the ethical way, this is the only legal approach. As a broker, you can't frontrun, etc.
I don't feel like I remember _Flash Boys_ making a case that GS was frontrunning its clients. In fact, the book I read was trying to make a case that the giant banks (and their clients) were victims of small shady HFT firms.
Nobody I've ever talked to that has worked in trading thinks _Flash Boys_ is a good book. There's a whole other book written as a rebuttal to it that is chock full of the technical details that Lewis' book lacks:
--- and, if you haven't read it (do! it's great!), know that the objections aren't about the morality tale Lewis is trying to tell, but rather that none of the details of his case against HFT actually make any sense.
Another good book that is recommended constantly in these threads is _Dark Pools_. Dark Pools is also a great sort of rags-to-riches story about the emergence of the most important electronic markets from tiny startups; it's a good read.
Interesting. I know Lewis' book about the industry I work in was, uh, an incomplete picture of things, so I'm not shocked if Flash Boys was a bit shaded too. Is there a shorter version of the objections to Flash Boys I can skim through to see if I curious enough to go further with this?
It's more than shaded in this case; it's literally incoherent. I think I can do a thematic tl;dr:
For the major commercial brokerages (at big investment banks, like RBC), it is literally their whole job to move large blocks of tradable instruments without moving the market in the process. Shopping large blocks is supposed to be difficult, because things are priced by the law of supply and demand.
If you're trying to sell (in RBC's case, for a client) 1000 lots of XYZ, you have an extremely important piece of insider information about XYZ: there are about to be 100,000 new XYZ shares on the market. Your attempt to sell all 100,000 is going to reduce the price of XYZ. Naturally, you want to capture the highest possible price for XYZ, so, naturally, you want to conceal your intent to sell 100,000 shares all at once.
Meanwhile, the market wants to do the best job possible of capturing all available information about XYZ and building into the current market price. Anything you do to tip your hand about your giant order will be analyzed by some market participant and used to execute a profitable trade (in this case, profiting by placing trades based on the [accurate] prediction that the price is going to fall).
The whole of _Flash Boys_ considers this trade a form of frontrunning, despite the fact that the traders doing it have no fiduciary responsibility either to RBC or to RBC's clients, and despite the fact that RBC's desire to sell at high prices involves RBC themselves exploiting an information asymmetry. In Lewis' world, the market is supposed to sit back and transact 1000 lots of XYZ without adjusting the price until after the giant corporations making that trade are done. It follows logically from Lewis' argument that market-making is itself somehow unethical.
This seems like minutiae, but (at least for the portions of the book about "level playing fields" and electronic markets) the theme running through the book is that markets should behave themselves and continue working in a way that matches the expectations of the equities desk at RBC. That's silly, but fine, until Lewis extrapolates from that the idea that these broken expectations harm Main Street investors. They don't, except to the extent that those investors predicate their returns on the largesse of big companies like RBC. The Chief Investment Officer at (unassailably awesome) Vanguard is on the record repeatedly about how electronic trading and HFT have improved returns for value investors and passively-managed funds.
Usual posts by rayiner tend to be rather conservative, and oftentimes, rather pro-government.
Seeing him arguing for something rather out of character is the joke. If you go back and look at his post history, you'll see some things going either way--but I don't think that makes the joke any less funny.
We should be a little more careful about assuming we can project entire ideologies onto people from their comments. I think it's probably pretty apparent why I'm sore about that point. Sorry to make a thing out of it though.
I think it's fine to model people's ideologies according to their comments, so long as we are able to adapt that model when new data happens. Similarly, if we consistently comment in ways that don't match our ideology, we shouldn't be upset when people draw incorrect conclusions about us.
Anyways, no harm done...it's just a few pg funbucks.
It's fine until you presume something that a reasonable person might find offensive. For instance: it bugs me when people extrapolate from my nerdy desire to close-read 4A law the belief that I generally support dragnet surveillance. It would also bother the heck out of me if someone presumed that I was in favor of onerous custodial sentences for most non-violent crime.
I agree mostly, but the exception is financial harm, which can devastate people and families. That retirement savings you were planning on using? Gone. When that happens, either restitution needs to be made or you go to jail.
That scenario is exactly why we need less jail time and more restorative justice [0] so the highest priority of the courts would be making sure the victims and community are compensated, any destruction is restored, and any emotional harm is healed. Rather than a strange obsession by the state to seek vindictive punishment, regardless of the interests of the victim or community - as is the current motivation in most countries.
Obviously this is an exaggeration but GS essentially played against their own clients in many cases. Edit: (There are stuff on the net about this, like http://www.huffingtonpost.com/2010/04/26/goldman-sachs-email..., but many of us know this system from the inside :) ).
Value-through-exclusivity doesn't mean that the entity itself is rivalrous.
Take my car, and now you have my car and I don't.
Take my trade secret, and now both of us have it. I haven't lost the knowledge itself.
The value may well have declined, however.
That very point was made by Thomas Jefferson on patents:
"If nature has made any one thing less susceptible than all others of exclusive property, it is the action of the thinking power called an idea, which an individual may exclusively possess as long as he keeps it to himself; but the moment it is divulged, it forces itself into the possession of every one, and the receiver cannot dispossess himself of it. Its peculiar character, too, is that no one possesses the less, because every other possesses the whole of it. He who receives an idea from me, receives instruction himself without lessening mine; as he who lights his taper at mine, receives light without darkening me."
A fair point, though there's more to it than that.
There are a lot of things which go into creating value for a company. Trade secrets are one of many factors. Execution, connections, business relationships, location (or locations for retail establishments), and more, all come into play.
Taking two well-known examples, the real value for both Coke and KFC is far less in the specific product forumulations of both companies -- those are effectively McGuffins. Rather it's in Coke's set of contracts with bottlers, distributers, retailers, and restaurants (many of which will carry _either_ Coke, _or_ Pepsi, but not both), etc. There are even "open source" colas whose recipies are published, but which don't have the market reach of the majors. For KFC, it's more a matter of retail outlets and supply chains -- having a global branding campaign, advertising and marketing, real estate in specified locations, and relationships with suppliers such that ingredients and product are supplied to these.
Yes, there are companies which do rely to a much greater extent on trade secrets, and there's some case to be made for some level of protection over these, particularly in tech. Michael Wolf's "Burn Rate" or ... I'm trying to remember the author and book, about the "Go" PDA who had their idea arguably poached by Microsoft. But in truth, execution and capitalization matter a tremendous amount, and exclusive knowledge of a specific concept, not so much.
I agree with this view of most intellectual "property", but it doesn't apply here. In the case of finance, trade secrets absolutely are rivalrous.
Strategies have limited capacities, so if anyone else is trading using your strategy you're going to be making less money. This is exacerbated by how markets are largely anonymous and distributed: nobody is trading with a given firm specifically, so somebody using your long-maintained strategy doesn't face any disadvantages against you.
Moreover, being secret is the only protection strategies can meaningfully have. Copyright can't cover the strategy itself, just the code to implement it, and patents by definition require divulging how it works. And even if you could somehow enforce a patent on a trading strategy, other people would still be able to trade against it, making the whole idea a non-starter.
At the same time I still don't think prison is a good answer to this sort of crime—but I don't think it's a good answer to car theft or most other sorts of crime either. But that's a function of deep issues I have with the justice system, a story best left for another time.
He will have willfully broken the terms of his contract. That's generally considered illegal and GS should have a fairly strong civil case against him given what the defence has admitted to.
Can't for the life of me think why this should be a criminal case though. Not sure how the American people are served by a conviction.
It is still technically in violation of a (civil) law so it could be considered illegal. But agreed, typically illegal is most often used in the context of criminal matters.
I have the strong sense that the New York court system, both federal and state, are particularly corrupt. The NYC based bankers all look out for each other. If you're plugged into the financial scene at the right levels in the NYC area, you're untouchable, and anyone who affronts you is going down.
How else can you explain John Corzine is still walking free and Madoff was never arrested for years despite many people screaming there was an obvious fraud? Why do you think they had to extradite some guy all the way from London in their search for a scapegoat for the Flash Crash? These people have a mutual protection gang going.
How else can you explain John Corzine is still walking free
I'm sure you know this, but for others not following so closely:
the commodities broker for the "guy" from
London was MF Global, the sleazy firm that
Corzine was running during the Flash Crash
So now they are trying to extradite a minnow, but the scum heading MF Global have so far avoided any fallout. As Wiki puts it[1]:
"Crimes have been committed here without a doubt.
We think there are enough facts out here to start
arresting people and start filing charges."
However, by August 2012, criminal investigators
had concluded that charges against Corzine,
or any other of MF Global's former executives
or employees would be unlikely.
MF Global allowed some clown trading out of his house to place orders for billions of dollars of futures contracts. If any reasonable percentage of those had been filled, the disruption to the markets would have been enormous. Not only did the "guy" in London not have enough capital to place those orders, they were big enough that MF Global itself didn't have enough capital. Of course, MF Global shortly later turned out to be a shell game, so maybe all that would have happened is they would have blown up a few months earlier.
This is totally ridiculous. The man is a veteran software architect. When you're a veteran software architect, you don't NEED the source code. The only thing it tells you is a few syntactical idiosyncrasies of whatever language/framework you happened to implement it in, and it saves you time in typing it out.
Ask anyone who is familiar with GS's code base whether they would rather start fresh or have GS's code base. I've done this, and you can tell from my tone what they said.
I've also been on the other side of such a case; someone copied some code I'd written. Guess what, they couldn't get it to work because they didn't understand how it worked.
I suspect the jury is composed of people who are easily talked into a "cooking recipe" view of how software works. Well, just as I'm not going to be a Michelin chef from copying a recipe, you are also not going to know how to design an HFT system if you don't have a deeper understanding of how it works. Or maybe more to the point: there are plenty of people who copy recipes and make tasty food (my parents are quite good at this, actually). They don't get paid what a Michelin chef gets paid, because a Michelin chef is not a guy who just mixes some ingredients according to a formula.
>Ask anyone who is familiar with GS's code base whether they would rather start fresh or have GS's code base. I've done this, and you can tell from my tone what they said.
I think that's a false dichotomy -- no one can remember every minute detail of every problem solved by a large application. To re-do the research would be a large commitment, hence I'm convinced the wisest choice would be "both," legality aside.
If we use your logic that it's hard to use other people's code, then the entire premise of open source is a failure. Clearly that isn't the case so I question your logic.
How timely... two-three years ago I had trudged through a pretty dated codebase to track down all the tables that contained relevant data to implement a new search system. Since then features were added, bolted on, and changes made (I had left the company for a year and a half in the middle).
Now, I'm migrating said data into a green solution that will replace the whole thing. It took a surprising amount of time to track it all down again, even with my old code and additional data. The new solutions data structures are a bit different than even the old search-only solution.
Even when you have the code, and even when you wrote it, that doesn't mean it's going to save you all the time... it's a little bit of a shortcut and reference, sure, but imho even if it's technically a crime, in this case I doubt it's enough of an offense to justify more than a month in jail, which has probably already been served.
A lot of the high frequency trading code actually is very expensive to recreate, because it contains logic like (and I am paraphrasing here):
if(bookPressure > 0.7 and alpha23signal<0.2)
buyEurUsd()
where constants like 0.7 and 0.2 were arrived at from looking at actual trade activity (and order book fill rates) conducted by the firm over a period of a few months. That can be incredibly expensive and or impossible to "recreate" from scratch.
The exact original code is also useful if you wanted to write code to "detect" that some automated trading is being done by a Goldmans proprietary algorithm, and then anticipate what it's going to do next so you can trade ahead of it.
Right but I don't see how this detracts from my point. Also, often the models are very simple - it's the calibration to a particular market that is very expensive (the constants - which are then often hard coded for best performance). This is because calibrating it requires access to data from actual trading activity (not just tick data, I mean also information on say the chance that your order will be filled in a particular situation: information you can only get by actually placing orders of size in the market and observing when they got filled).
If you don't know how that constant is arrived at, how do you know whether it still applies to the current market? Don't you want to know whether your assumptions still hold? And aren't you stealing the wrong thing, then?
If you do know, why do you care what the number actually is?
And why would hard coding improve performance? Compiler optimization?
This is an interesting point I hadn't seen made before. If it applies here, then what you're saying is that the structure of the code itself doesn't matter, so much as the parameters encoded into it --- those parameters being available only to people who can bankroll the trading desks that generate the information from which they're derived.
In that scenario, the "code" itself (as we understand "code") is a red herring, right?
I'm surprised if this is what actually happened; it's too reasonable-sounding.
I would hope that I would encounter source code that at least took me to the code that derived these magic numbers. Why 0.7 for bookPressure or 0.2 for alpha23signal?
The fact that the Manhattan DA's office spent so much effort pursuing this case goes to show how much power GS has over there. They are trying to make an example out of him. The state is literally doing GS's bidding here. Such a shame.
Agreed. While it's important for law enforcement to step up to new and technically challenging cases, the public would have been much better served had the DA focused on the Dewey & LeBoeuf case (as the NY Times suggests). This case had small marginal negative impact on society (and the leaking of HFT technology might even result in a positive for investors). Dewey & LeBoeuf are outright accounting fraudsters. Great story: http://www.newyorker.com/magazine/2013/10/14/the-collapse-2
"The prospect of a mistrial stemmed from a dispute between two jurors deciding Mr. Aleynikov’s fate, with a female juror accusing a male one of “food tampering” — in part because an avocado was missing from her sandwich."
This has been discussed over and over, but organisations aren't required to release GPL code that never leaves their servers. GPL code belongs to the organisation and the organisation alone if hired programmers work on it on behalf of the organisation.
The programmer in question did not own the code, GS did. By releasing the code, he's stealing from them. End of story.
Source - I've actually read the whole GPL licence.
Indeed, it has been discussed many times. But I don't remember any discussion on attribution rights. If GS replaced GPL text with their own header they presumably deleted names of the original authors. I wonder if those authors could (collectively perhaps) send a complaint to GS and ask to restore the attributions?
To be pedantic, the GPL doesn't actually say anything about the contract between an employer and employee. While I highly doubt the Goldman Sachs employment contract is anything other than work-for-hire, meaning you're still right, it's not impossible for a programmer to retain copyright of code they write (eg, any code you write on your own time on a computer you own).
I am quite grossed out by the racist and misogynist media coverage of the "missing avocado" incident. Every single article about the case makes sure to mention that it was a female juror against a male juror and some of the articles also point out that the accuser is asian while the accused is white. How are these factoids relevant to the incident, or the larger case?
107 comments
[ 1.9 ms ] story [ 202 ms ] thread> Mr. Marino did not challenge the underlying facts of the case. Instead, he said they added up not to a crime but to a simple breach of Goldman’s confidentiality policy.
> “There’s no doubt Mr. Aleynikov did something wrong,” he told the jury, stopping occasionally to sip from a water bottle he clutched to throughout the closing statement. “But this is a very badly misguided case.”
There has been a lot of misinformation surrounding this case. I think its well established now that Sergey Aleynikov did indeed take more source code than just his open source contributions with him, his lawyer admits as much.
So I guess the take away is on your last day of work, don't just zip up and ftp any code from your company, it's just not worth it.
Actually I think the real takeaway is that if you are really good at your job, your company will pay more attention to you, especially when you leave. Sergey was a big part of the HFT team at Goldman, he literally rebuilt their infrastructure and his leaving was a huge deal to the team, and he wasn't just leaving to go to say Google, he was going to a new HFT outfit that would compete with Goldman.
His actions would be like Mark Lucovsky leaving Microsoft with some windows NT kernel code he wrote, or Jeff Dean leaving Google with parts of the deep learning code to go to competitors to work on the same thing. I'd be willing to bet that both google and Microsoft would get lawyers involved in both those cases.
If you want to make open source contributions while at work join a company like, um...., Microsoft?
EDIT toned down the language calling him the entire HFT talent at Goldman, it was a bit over the top:)
Not just "not worth it" but actually illegal.
The only question of debate here is whether jail time was warranted here.
They only mention it in passing, but the reason he got all that evidence thrown out was essentially that the police were acting on very little else than Goldmans assertion.
Then he got acquitted on what was if you want to a "legal loophole", but still (and even that only happened a few steps up on the jurisdictional ladder, even though the law clearly didn't apply to him).
And then they charged him again, with essentially the same crime. So the same people that can't manage to put a single Goldman dude behind bars are all too happy to destroy lives on Goldmans words.
One bias prosecutors definitely have is taking cases they think they can win.
That's a fair statement by the OP. The external criticism has always been there was a lack of a full investigation from a neutral position by the state, as investigators depended on Goldman Sachs people to determine the crime.
Outside neutral parties weren't consulted AFAIK. Which is typically the case when sophisticated crimes take place - such as bringing in private industry experts.
Whether or not that would have resulted in different charges or handling by the police is up for debate. But it is certain that the worst of the original charges were found to be baseless, prosecutors were over-aggressive, and most of the evidence was thrown out because they didn't have a proper basis to conduct search warrants.
Much in the same way that ICE has acted like the enforcement arm of the RIAA/MPAA when taking down domains. Barely (or not at all) verifying that the domains were being used for infringement (instead just taking a list given to them by the RIAA/MPAA and acting on it blindly).
I agree with the prevailing sentiment on HN that (a) it's too easy to get an indictment (the grand jury is supposed to be adversarial against the prosecutors and isn't) and (b) prosecutors bias --- my belief is: primarily bias towards cases they can win easily --- has subverted the goal of prosecutorial discretion.
But the facts of this case suggest a theft that could reasonably be valued at extremely high numbers. For it to be heard in court doesn't sound like a miscarriage of justice.
They would have probably already dropped this case the moment most of the evidence was ruled out and called it a day.
* was sentenced
What do you have to say to that?
Michael Lewis wrote a big article in Vanity Fair which I suggest you look at. He gathered a number of experts who looked at the evidence presented at his federal trial and concluded that what he took was modified open source code to run the queues, do messaging, etc. Boiler plate code to create the infrastructure.
This is what GS is calling "proprietary source code", since he made the changes on GS dime therefore it is proprietary to them.
In addition to proprietary source code, Aleynikov also transferred some open source software licensed for use by the public that was mixed in with Goldman’s proprietary code. However, a substantially greater number of the uploaded files contained proprietary code than had open source software.
I encourage you to look at Michael Lewis's article[1]. This passage is relevant: "As Goldman hadn’t permitted him to release his debugged or improved code back to the public—possibly in violation of the original free licenses, which often stated that improvements must be publicly shared—the only way to get his hands on these was to take the Goldman code. That he had taken, in the bargain, some code that wasn’t open source, which happened to be contained in the same files as the open-source code, surprised no one. Grabbing a bunch of files that contained both open-source and non-open-source code was an efficient, quick, and dirty way to collect the open-source code, even if the open-source code was the only part that interested him."
[1] http://www.vanityfair.com/news/2013/09/michael-lewis-goldman...
As Aleynikov concedes in his motion papers, the code he took from Goldman Sachs included a "purposefully designed" portion of the Goldman Sachs "proprietary, custom-built trading system." Indeed, the evidence showed that Aleynikov took a significant percentage of the proprietary source code for that system. While Aleynikov attempted to show that there was open source code embedded within the proprietary code and to identify the files in which that might be true, his expert witness was only able to identify one file among those taken by Aleynikov that both bore a Goldman Sachs copyright banner and appeared to contain open source code.
Sure, the modified open-source code might have been protected by GS employment agreement but that doesn't mean that it was their secret sauce. It was infrastructure/boiler plate code.
Additionally the fact that a code has a GS copyright banner doesn't mean that it was all brand-spanking-new GS-unique code. Even Lewis's article talks about the fact that a single file had open-source code and modified open-source code, aka proprietary code. This subtle language difference is exactly how a lawyer confuses the jury and it is difficult to counteract. Remember the SCO Unix v Novell fiasco!?
Lewis brought together experts who reviewed the evidence on record, i.e. the actual source codes submitted at evidence at his first trial and they reviewed it and found it to be derivative open-source code to run queues, and run infrastructure.
Your source, by the way, does not support your assertion about Lewis contracting experts reviewing the source code presented as evidence during the trial. Which makes sense: what would make you think Lewis would have access to it in the first place?
No one from GS has said that what he has taken was an algorithm or anything related to decision making. Not one person.
I never said the Second Circuit resolved a dispute about it being proprietary code. The defendant's own lawyer doesn't dispute that it was proprietary. But merely being proprietary is not sufficient for a trade secret violation. If you modify a config file for Apache that modified config file becomes proprietary but it is most likely not an essential trade secret that is the core of your business that someone else couldn't have come up with something similar to solve your problem.
GS lawyer's say that proprietary code WAS trade secret and essential, whereas defendant says no it was modified open source code used for queueing etc. The jury is not sophisticated enough to differentiate between the two. See any patent case where a BS software invention is held up as unique.
One of the legal qualification of trade secrets is that it is a method, process, etc. that isn't known or reasonably known by others for gaining advantage over competitors. It is their secret sauce.
I don't see anything here that suggests he took the ALGORITHM or DECISION MAKING process, or any such thing.
Like working for Goldman Sachs?
https://en.wikipedia.org/wiki/Goldman_Sachs#Controversies
Among the HN crowd I am sure you will find people on both sides of this issue: some people will think that GS is not working on 'good' problems that better the world, therefore, it is 'shady' to work there; and others will argue that GS is abstractly bettering the world through efficient resource/money allocation, especially for developing countries.
This would be true if GS didn't scam people, which they do. I accept the investment allocation argument, but I'm not convinced that it's valid for GS specifically.
EDIT: Holy shit, that controversies section just goes on and on. I just checked, and it's 7000 words long. This company's insane.
That's what I was thinking. That's an awful lot of smoke if there's no fire.
Nonsense. Lloyd Blankfein, the CEO of Goldman, is on record that the company is just "doing God’s work". http://dealbook.nytimes.com/2009/11/09/goldman-chief-says-he...
/sarcasm
This is not just the ethical way, this is the only legal approach. As a broker, you can't frontrun, etc.
http://www.amazon.com/Flash-Boys-Insiders-Perspective-High-F...
--- and, if you haven't read it (do! it's great!), know that the objections aren't about the morality tale Lewis is trying to tell, but rather that none of the details of his case against HFT actually make any sense.
Another good book that is recommended constantly in these threads is _Dark Pools_. Dark Pools is also a great sort of rags-to-riches story about the emergence of the most important electronic markets from tiny startups; it's a good read.
For the major commercial brokerages (at big investment banks, like RBC), it is literally their whole job to move large blocks of tradable instruments without moving the market in the process. Shopping large blocks is supposed to be difficult, because things are priced by the law of supply and demand.
If you're trying to sell (in RBC's case, for a client) 1000 lots of XYZ, you have an extremely important piece of insider information about XYZ: there are about to be 100,000 new XYZ shares on the market. Your attempt to sell all 100,000 is going to reduce the price of XYZ. Naturally, you want to capture the highest possible price for XYZ, so, naturally, you want to conceal your intent to sell 100,000 shares all at once.
Meanwhile, the market wants to do the best job possible of capturing all available information about XYZ and building into the current market price. Anything you do to tip your hand about your giant order will be analyzed by some market participant and used to execute a profitable trade (in this case, profiting by placing trades based on the [accurate] prediction that the price is going to fall).
The whole of _Flash Boys_ considers this trade a form of frontrunning, despite the fact that the traders doing it have no fiduciary responsibility either to RBC or to RBC's clients, and despite the fact that RBC's desire to sell at high prices involves RBC themselves exploiting an information asymmetry. In Lewis' world, the market is supposed to sit back and transact 1000 lots of XYZ without adjusting the price until after the giant corporations making that trade are done. It follows logically from Lewis' argument that market-making is itself somehow unethical.
This seems like minutiae, but (at least for the portions of the book about "level playing fields" and electronic markets) the theme running through the book is that markets should behave themselves and continue working in a way that matches the expectations of the equities desk at RBC. That's silly, but fine, until Lewis extrapolates from that the idea that these broken expectations harm Main Street investors. They don't, except to the extent that those investors predicate their returns on the largesse of big companies like RBC. The Chief Investment Officer at (unassailably awesome) Vanguard is on the record repeatedly about how electronic trading and HFT have improved returns for value investors and passively-managed funds.
anything we can do?
Seeing him arguing for something rather out of character is the joke. If you go back and look at his post history, you'll see some things going either way--but I don't think that makes the joke any less funny.
I use different things when poking fun at you. :)
Anyways, no harm done...it's just a few pg funbucks.
[0] https://en.wikipedia.org/wiki/Restorative_justice
Obviously this is an exaggeration but GS essentially played against their own clients in many cases. Edit: (There are stuff on the net about this, like http://www.huffingtonpost.com/2010/04/26/goldman-sachs-email..., but many of us know this system from the inside :) ).
Take my car, and now you have my car and I don't.
Take my trade secret, and now both of us have it. I haven't lost the knowledge itself.
The value may well have declined, however.
That very point was made by Thomas Jefferson on patents:
"If nature has made any one thing less susceptible than all others of exclusive property, it is the action of the thinking power called an idea, which an individual may exclusively possess as long as he keeps it to himself; but the moment it is divulged, it forces itself into the possession of every one, and the receiver cannot dispossess himself of it. Its peculiar character, too, is that no one possesses the less, because every other possesses the whole of it. He who receives an idea from me, receives instruction himself without lessening mine; as he who lights his taper at mine, receives light without darkening me."
http://www.businessinsider.com/thomas-jefferson-on-patents-2...
There are a lot of things which go into creating value for a company. Trade secrets are one of many factors. Execution, connections, business relationships, location (or locations for retail establishments), and more, all come into play.
Taking two well-known examples, the real value for both Coke and KFC is far less in the specific product forumulations of both companies -- those are effectively McGuffins. Rather it's in Coke's set of contracts with bottlers, distributers, retailers, and restaurants (many of which will carry _either_ Coke, _or_ Pepsi, but not both), etc. There are even "open source" colas whose recipies are published, but which don't have the market reach of the majors. For KFC, it's more a matter of retail outlets and supply chains -- having a global branding campaign, advertising and marketing, real estate in specified locations, and relationships with suppliers such that ingredients and product are supplied to these.
Yes, there are companies which do rely to a much greater extent on trade secrets, and there's some case to be made for some level of protection over these, particularly in tech. Michael Wolf's "Burn Rate" or ... I'm trying to remember the author and book, about the "Go" PDA who had their idea arguably poached by Microsoft. But in truth, execution and capitalization matter a tremendous amount, and exclusive knowledge of a specific concept, not so much.
Strategies have limited capacities, so if anyone else is trading using your strategy you're going to be making less money. This is exacerbated by how markets are largely anonymous and distributed: nobody is trading with a given firm specifically, so somebody using your long-maintained strategy doesn't face any disadvantages against you.
Moreover, being secret is the only protection strategies can meaningfully have. Copyright can't cover the strategy itself, just the code to implement it, and patents by definition require divulging how it works. And even if you could somehow enforce a patent on a trading strategy, other people would still be able to trade against it, making the whole idea a non-starter.
At the same time I still don't think prison is a good answer to this sort of crime—but I don't think it's a good answer to car theft or most other sorts of crime either. But that's a function of deep issues I have with the justice system, a story best left for another time.
Can't for the life of me think why this should be a criminal case though. Not sure how the American people are served by a conviction.
But I'm not sure I would call it "illegal".
How else can you explain John Corzine is still walking free and Madoff was never arrested for years despite many people screaming there was an obvious fraud? Why do you think they had to extradite some guy all the way from London in their search for a scapegoat for the Flash Crash? These people have a mutual protection gang going.
I'm sure you know this, but for others not following so closely:
So now they are trying to extradite a minnow, but the scum heading MF Global have so far avoided any fallout. As Wiki puts it[1]: MF Global allowed some clown trading out of his house to place orders for billions of dollars of futures contracts. If any reasonable percentage of those had been filled, the disruption to the markets would have been enormous. Not only did the "guy" in London not have enough capital to place those orders, they were big enough that MF Global itself didn't have enough capital. Of course, MF Global shortly later turned out to be a shell game, so maybe all that would have happened is they would have blown up a few months earlier.[1] https://en.wikipedia.org/wiki/MF_Global#April_2012:_Possible...
Ask anyone who is familiar with GS's code base whether they would rather start fresh or have GS's code base. I've done this, and you can tell from my tone what they said.
I've also been on the other side of such a case; someone copied some code I'd written. Guess what, they couldn't get it to work because they didn't understand how it worked.
I suspect the jury is composed of people who are easily talked into a "cooking recipe" view of how software works. Well, just as I'm not going to be a Michelin chef from copying a recipe, you are also not going to know how to design an HFT system if you don't have a deeper understanding of how it works. Or maybe more to the point: there are plenty of people who copy recipes and make tasty food (my parents are quite good at this, actually). They don't get paid what a Michelin chef gets paid, because a Michelin chef is not a guy who just mixes some ingredients according to a formula.
I think that's a false dichotomy -- no one can remember every minute detail of every problem solved by a large application. To re-do the research would be a large commitment, hence I'm convinced the wisest choice would be "both," legality aside.
Cooking and programming are horrible comparisons too.
I think that was part of their point
This isn't necessarily a good sign.
Now, I'm migrating said data into a green solution that will replace the whole thing. It took a surprising amount of time to track it all down again, even with my old code and additional data. The new solutions data structures are a bit different than even the old search-only solution.
Even when you have the code, and even when you wrote it, that doesn't mean it's going to save you all the time... it's a little bit of a shortcut and reference, sure, but imho even if it's technically a crime, in this case I doubt it's enough of an offense to justify more than a month in jail, which has probably already been served.
if(bookPressure > 0.7 and alpha23signal<0.2) buyEurUsd()
where constants like 0.7 and 0.2 were arrived at from looking at actual trade activity (and order book fill rates) conducted by the firm over a period of a few months. That can be incredibly expensive and or impossible to "recreate" from scratch.
The exact original code is also useful if you wanted to write code to "detect" that some automated trading is being done by a Goldmans proprietary algorithm, and then anticipate what it's going to do next so you can trade ahead of it.
In which case you need to understand the model.
This caught my eye. Really?
If you do know, why do you care what the number actually is?
And why would hard coding improve performance? Compiler optimization?
In that scenario, the "code" itself (as we understand "code") is a red herring, right?
I'm surprised if this is what actually happened; it's too reasonable-sounding.
I would hope that I would encounter source code that at least took me to the code that derived these magic numbers. Why 0.7 for bookPressure or 0.2 for alpha23signal?
The justice system in action.
The programmer in question did not own the code, GS did. By releasing the code, he's stealing from them. End of story.
Source - I've actually read the whole GPL licence.