Oracle copied Amazon’s API – was that copyright infringement?
arstechnica.com
arstechnica.com
> Did Oracle infringe Amazon's copyright here? Ars Technica contacted Oracle to ask them if they had a license to copy Amazon's S3 API. An Oracle spokeswoman said that the S3 API was licensed under an Apache 2.0 license. She pointed us to the Amazon SDK for Java, which does indeed come with an Apache 2.0 license.
> However, the Amazon SDK is code that uses the S3 API, not code that implements it—the difference between a customer who orders hash browns and the Waffle House cook who interprets the orders.
While I'm not a lawyer and didn't study the Google case in that much detail, I don't see how they can claim that their own Java APIs can be copyrighted, and then defending their copying of Amazon's API on the grounds that an SDK using it is open source. I would think they could just look at the documentation, and skip the SDK altogether. :)
Wall garden, vendor lock, license, and litigate for every LOC. Make customers pay for technical debt by dumping more exotic spaghetti code.
Larry Ellison's way of doing business is about him and what he receives at the bottom line. The customer is an after fact in the sales funnel. A resource that needs to be milked as a proper cash cow.
Larry Ellison is as a evil as the offspring of Joffrey and Cersei Lannister could be.
What would be a better option (IMO) then just not enforcing your rights would be to get an nominal fee from the entities you like and in return grant them a favourable license.
(Not a lawyer, go ask your own lawyer if you want specific legal advice)
For an example, lets just say for the sake of argument Amazon did decide to defend their API from Oracle's implementation and take action. Oracle could use as a defence (But prob not their only line of defense) that "That Internet Archive, LeoFS, Riak CS, Minio, cloudian and others all have S3-Like API's. Amazon must be aware that these products exist and by allowing such products to exist without defending their rights (what ever they are) to the API they allowed an implied license for the API to be created thus Oracle's implementation is just another implementation of that implied license.
A "Workaround" to such could be taking lets say Internet Archive to one side and saying "Hey, we like what you do, but we have a copyright on that API you are using. Pay us $1 and we will grant you a license to use that API.".
Personally I dislike the idea of enforcing copyrights on an API (Sure, enforce them code behind that API, unless you wish to give that away) and feel that the products using them should stand on their own.
Implied license generally applies to situations where a reasonable person would believe they have a license even though one has not been explicitly granted. For example, if I pay you for the use of a photograph you took, even without a written contract granting me ownership or a license, a court would hold that I have an implied license.
Failing to defend IP can enable a laches defense, but that's quite different from an implied license.
Off the top of my head (Which may not be a great example but the first one that sprung to mind) Field v. Google.
> The Field court expanded the earlier doctrine to hold that implied license arises "where the copyright holder knows of the use and encourages it." Applying this two-pronged knowledge and encouragement test to the Field facts, the court found that the website owner was aware of the robots exclusion protocol mechanisms for communicating with the Googlebot, and that by not using them, he essentially encouraged the Googlebot to index his content.
> Field created a robots.txt file for his site and set the permissions within this file to allow all robots to visit and index all of the pages on the site. [...] Field created the robots.txt file because he wanted search engines to visit his site and include the site within their search results.
and (regarding meta-tags):
> Field concedes he was aware of these industry standard mechanisms, and knew that the presence of a “no archive” meta-tag on the pages of his Web site would have informed Google not to display “Cached” links to his pages. Despite this knowledge, Field chose not to include the no-archive meta-tag on the pages of his site. He did so, knowing that Google would interpret the absence of the meta-tag as permission to allow access to the pages via “Cached” links. Thus, with knowledge of how Google would use the copyrighted works he placed on those pages, and with knowledge that he could prevent such use, Field instead made a conscious decision to permit it. His conduct is reasonably interpreted as the grant of a license to Google for that use.
Field was trying to get his site indexed and cached (so that he could sue Google), and Google indexed and cached it. The court ruled that Field's actions were active encouragement and implied the grant of a license to do the thing he wanted Google to do.
Field didn't fail to defend his copyright. He put the files up in January, filed copyrights with the Copyright office, waited for them to be indexed and cached, and sued Google in April. He protected his copyright about as vigorously as you can with the available legal tools.
The question of implied license in this case is whether putting something on the internet without taking steps to stop search engines from indexing+caching it constitutes an implied license. The court ruled it did, but it relied in part on Field taking specific actions to allow it (such as the robots.txt file), so it's not entirely clear that in a different case they would rule the same way. If someone put up a site with just a plain text file without HTML tags and no robots.txt file maybe the court would rule that there was not an implied license there, I don't know. They also took into consideration that there are well-known industry standard ways of preventing Google and others from indexing and caching your site, so maybe they would consider that alone sufficient to rule that there's an implied license if you don't follow the industry standards.
In any case, Field v. Google has little-to-nothing to do with failing to defend a copyright.
As a legal strategy, that's not a problem, and in fact it's what lawyers are supposed to do: make the best possible arguments they can in any case they are a part of. It should not be interpreted as a case of corporate schizophrenia -- it's just the lawyers doing their jobs.
This can lead to contradictory positions if multiple cases are going on at the same time. That's fine. The legal system is designed around the assumption that this will happen.
However, I think that Oracle would argue that there is no contradiction.
In the case now before the US Supreme Court, Oracle argues that software copyrights do extend to the interface/API and that reimplementing the interface does not constitute fair use.
With regard to Amazon's API, Oracle argues that the license grants them permission to reimplement the interface.
This argument doesn't rely on disputing copyrightability of APIs or on fair use. It simply says that copyright holders get to decide whether or not reimplementation is permitted.
Fact in this matter is that the license for the Amazon API does allow the API to be copied. But Oracle did and does not follow the license, because the required attribution is missing. They are therefor not protected by the license. It is like using an illegal copy of Windows and claiming you had the right to use it under the license for the small omission of not having paid for it.
Even if the change the code to include attribution they are still liable for significant damages to Amazon for the period preceding the correction. Amazon can claim to them the attribution is a very significant part of the license agreement: It would deter competitors having to advertise Amazon and tell their customers that their product was actually invented by someone having a competing product.
Google can claim their fair use was actually already common practice, as most of us thought. Even Oracle itself assumed and acted on that assumption that copying APIs was fair use / not copyrightable and did not require following a license agreement, as they did not. Worst case for Google is that they should not have to pay significant damages for something they and others would see as a common mistake and Oracle's reversal complying with the Amazon license is proof that there was no established judicial practice to follow.
So Oracle is in a pickle: Either they claim copying the Amazon API was fair-use, or APIs are not copyrightable and they lose the case against Google. Or they claim that the license of Amazon allowed for copying of the API and they violated that license. The harder they push Google in the later case the bigger the risk Amazon will claim their share of Oracle's pie in return. Better get the spreadsheets out.
Also, the potential damages in these two cases are presumably very different. Google built an entire industry on top of Android whilst Oracle is limping along with a me-too cloud offering, some tiny part of which uses Amazon's API. So I think Oracle would gladly take the risk of losing that case if they can win the other one.
If a company is filling a role and 2 similar applicants get matching offers and one accepts the first and the other negotiates I wouldn’t expect the company to go back to the first person and offer him more to bring them to parity.
Oracle's case with Google will likely set precedent. If the side they're arguing for is going to hurt another side of their business by removing their legal argument, why are they fighting?
They took an open source piece of software (the SDK) and made a derivative work (the backend the SDK uses). That indeed is copying and is essentially identical to what Google did. But there's no copyright violation here because Amazon explicitly granted everyone the right to do that by releasing the SDK under the Apache license.
I fail to see how it substantively differs.
Google acknowledge's this and isn't bothering to try and argue otherwise. They aren't presenting openjdk's license as a defense to infringement. Oracle for whatever reason is ignoring that issue in their statement to ars technica.
https://github.com/oracle/oci-ruby-sdk/blob/master/LICENSE.t...
When Oracle sued Google, the case originally had both copyright claims and patent claims.
Is Amazon's API patented? If not, then if Amazon were to sue Oracle it would be only a copyright case, not a patent and copyright case.
Normally, appeals from most cases, including copyright cases, go to the appeals court for the same Circuit that the district court that heard the case is in, which would be the 9th Circuit Court of Appeals (9th CA) in the case of Oracle vs. Google, and most likely also in an Amazon vs. Oracle.
However, if the case includes patent claims the appeal goes to the Court of Appeals for the Federal Circuit (CAFC) instead of going to the CA for the circuit containing the district court.
If the case includes both patent claims and non-patent claims, the whole thing goes to CAFC which decides both the patent issues and the non-patent issues.
On patent issues, CAFC decisions are precedent for all district courts.
On those issues that get dragged along when a mixed case ends up at CAFC because it includes patent issues, CAFC is supposed to follow the precedent of the appeals court for the Circuit the case came from.
A lot of people believe that CAFC totally botched following 9th CA precedent in the Google case.
If Amazon sues Oracle but doesn't have a patent claim included to force appeals to CAFC, the appeal will end up at the 9th CA, which presumably will do a better job of following 9th CA precedent than CAFC did, and so rule against the API copyright (and so in favor of Oracle).
Until the Supreme Court actually gets around to settling this matter, we are in an inconsistent state in the 9th Circuit.
1. As far as specifically Google and Oracle are concerned, Oracle's API is copyrighted and what Google did infringes. CAFC said so when it handled the appeal, and that is binding precedent to the lower courts in any further proceeding between those two litigants.
2. For API copyright claims that are not between Oracle and Google and are not accompanied by a patent claim, the law is still what it was before the Google case.
3. For API copyright claims that are not between Oracle and Google but are accompanied by a patent claim, and so will be appealed through CAFC instead of 9th CA, presumably they will end up with the same wrong outcome as Oracle vs. Google.
While that is indeed the difference between using the API and implementing the API, I was under the impression that the Google-Oracle dispute was about the API itself, not the implementation. In the restaurant analogy, that would be the menu ("hash browns"), not the cook ("the customer wants hash browns, what are you going to do about it?").
This is common any more in legal battles. Where a company will take a position in one case that is completely contrary to its position in a different legal case
There is no need for simple 'because they are evil'-comments. It doesn't help discussion.
A causal reader might not be aware of how hypocritical certain operators are in the IT space and are reading this for the first time.
Just because it's obvious TO YOU doesn't mean it's obvious to everyone.
Though I do agree that lots of "because they're evil" comments do not add to the discussion.
Personally I think downvoting should be done carefully and not because the downvoter disagrees with the comment.
I agree. I just forget occasionally.
Usually these comments are much more concise than whatever fluff article is linked. (Or they provide some required context for posts that are very much in need of it.)
Maybe it doesn't increase the level of discussion here, but maybe there's value in steering the scarce resource of intelligent conversants toward more productive threads? Additionally, getting people up to speed on corporate citizenship of various companies is an ongoing process, an ambient meta conversation that pervades many conversations.
'They just want money' is too simplistic even if that is the case, because can they afford to keep doing this to everyone? Is there a strategy? And if it becomes a bandwagon of people with simple explanations there will be resistance towards anyone who figures out what could be their reasoning.
The Oracle Amazon API usage is an identical re-implementation of the interfaces, but as a web solution there is per se nothing to copy in verbatim form.
The IBM-PC copycats back in the 80s went as far as clean room re-implementing the system with engineers who never saw the documentation. Ripping the header files out of a proprietary license package is quite cheap from a legal standpoint in comparison, even if open source is provided.
They also have Lots of Med to Large organizations using a number of their products, a lot of them are Oracle customers simply because they were using products that oracle acquired and since they are now entrenched into the business process it is easier to keep paying Oracle than extract those products with alternatives
I don't know what it would mean for someone to "own SQL". It's an ISO standard:
https://en.wikipedia.org/wiki/SQL#Interoperability_and_stand...
> https://www.cnet.com/news/sun-wins-java-iso-approval/
That article describes when Sun was recognized as an official submitter for a proposed standardization of Java. It later withdrew.
> "In 1997, Sun Microsystems approached the ISO/IEC JTC 1 standards body and later the Ecma International to formalize Java, but it soon withdrew from the process."
https://en.wikipedia.org/wiki/Java_(programming_language)#Hi...
Java is not an ISO standard.
At the very best it's extremely grey area in the law.
In fact, if Oracle is right, the ISO is likely violating the copyright of the original authors of SQL by even publishing that standard.
Sun set up a certification process for Java, but specifically had different rules for mobile applications, because that's where they were trying to make money. Not that I agree with Oracle's argument, but it was very willful on Google's part to avoid licensing.
tldr; SQL is not analogous to Android.
And of course no one thought to ask for a real license, because no one thought it was copyrightable.
Willful on Google's part to avoid licensing is sort of synonyms with saying Google willfully complied with the law... I'm not sure how it's relevant.
It definitely isn't relevant to the rest of this discussion... you're comparing the attitude of the producer of the allegedly copyrighted material in one case to the attitude of the infringer of the allegedly existing copyright in another.
The reason is to prevent cases where company first submits something to the standardization body and then tries to extort money from others trying to implement that standard.
This of course is not a foolproof mechanism. There could be for example companies not participating in the standardization who hold significant IPR in the field. Certainly the policies between standardization bodies differ. Not sure how the ISO/IEC handles these issues.
(More realistically though, it'd unfortunately mostly result in mass renaming to .destroy())
You can't copyright them either in the same way that Harry Potter is copyrighted but not the word 'the' which is used throughout.
But yes, I'm pretty sure they can get away with this in the sense of avoiding submitting themselves to the their interpretation of the law unless/until it's ruled correct by a court.
The Google-Oracle case looks like it will be heard by the supreme court sometime in March, we'll get a ruling before the court goes out of session for the summer (the court takes a summer break).
I'm pretty sure even if Amazon sued Oracle now, that court would be happy to stay the case until the Supreme court rules. Even if the court wasn't actually willing to stay the case, it seems highly unlikely it would be at all resolved before the supreme court rules. Oracle is capable of delaying for a few months...
IANAL
the amazon plea is different because amazon published their SDK AND gave it a permissive license. further oracle is not looking at the S3 code to figure out how to implement the API, they must implement the functionality of the API based off guesses of the client SDK.
The implementation of Java and the implementation of the AWS API does not matter, since the Oracle-Google suit is (at this point) strictly over the structure sequence and organization of the APIs. How the code implementing the API came to be is completely and wholly irrelevant to the suit at hand.
Not that it matters (as explained above), but you're also wrong that google took java and executed 's/java/GoogleJava/g'. They re-implemented the entire API in dispute by hand, the only portion they copied from Oracle's implementation was the API.
The real victory for humanity would be the end of copyright itself. Too much time is wasted on trying to understand and work around this fundamentally broken system.
If we keep relying on copyright, we'll eventually lose everything that makes computers great. Copyright infringement is trivial because we own our machines and have full control over the software that runs on them. Governments and the copyright industry are growing increasingly hostile to the idea that citizens should be able to run whatever software they want. Therefore, maintaining the illusion that information is scarce is fundamentally incompatible with the freedoms envisioned by Stallman's licenses, especially the latest versions.
This API copyright issue also threatens competing free software implementations of proprietary APIs. Projects like Wine should be directly affected by the court's decision.
Look, should a CSS snippet copywriteable? What about a complex SQL query?
Imo, no singular LOC is unique in the way that a newly developed drug is unique or a song is unique. If devs can agree on that, why should that extend to an arbitrarily long combination of LOCs?
Even if it's somehow copyrightable, why should the rights of that LOC extend to the person claiming theirs has been breached? It's a bold claim to say that you are the first dev in the world to have ever created a particular piece of functionality.
Code should be treated like trade secrets - the protection of the creator is in the meticulous seperation of end product from recipe. SaaSs already do this, the entire front end is publicly visible code, which line-for-line, rarely accounts for more than a small % of the backend logic.
That would also bring about an abrupt end of license wars of permissive vs. copyleft, I suppose.
They're copying the S3 API, and they did so under the license granted in the S3 client software.
If you publish either your client or server software, or some specification of your API and you do so with some kind of license, you are licensing your API.
Amazon's publishing of their client software under Apache means they have licensed their API to everyone who downloaded it, including Oracle. The license is irrevocable.
Google copied Java's API without getting license to do so. Java was not published by Oracle or Sun with such a license, nor did Google obtain such license to use their API by contacting Oracle for one.
The I in API means interface. Ars is getting confused by the API and the implementation of it here:
> However, the Amazon SDK is code that uses the S3 API, not code that implements it—the difference between a customer who orders hash browns and the Waffle House cook who interprets the orders. Code that uses an API will be organized completely differently from code that implements one; it may not even contain the whole API. And Oracle has for years argued that using an API is unrelated to reimplementation and not an infringement of copyright (or else every app developer using Java would infringe).
The client library isn't the code using the API, it contains the the API. Oracle argues in the link:
> This case is about Google’scopying of Oracle’s work. App programmers are unaffected. They remain free to write any program they want availing themselves of the Java APIs, for free. The decision only requires platform developers — usually large corporations with tremendous resources, like Google—to take a license if they want to copy Oracle’s work and incorporate it into a substitute commercial product.
Oracle's point here is that an application using a small portion of the API is acceptable copying. AWS published and licensed the entire client software, containing the entire API for S3.
This is plainly false. To do so under the license they would have had to comply with the terms of the license, which include attribution. They did not do so.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
[1] https://github.com/aws/aws-sdk-java-v2/blob/master/LICENSE.t...
https://docs.cloud.oracle.com/iaas/api/
Not to mention, they reproduced it on the server even if they didn't distribute it. Copyright protects making copies just as more (more really) as distributing them.
Patently false. Java was published by Sun with an open source license over a decade ago: https://en.wikipedia.org/wiki/OpenJDK
The whole Harmony TCK situation was a clusterfuck, because Sun was ostensibly for Free Software/ Open-Source, but a certified, Apache-licensed Java implementation would have killed the official JDK. Harmony was donated to the ASF by IBM - Sun's erstwhile rival. As an outsider, that era of Sun vs. IBM proxy wars was...interesting to follow. I had a chuckle when I learned that the "Eclipse" IDE was named to throw shade at Sun, and it was from a salty Sun CEO/employee.
It doesn't matter though, Google didn't comply with OpenJDK's GPL license (since they, like everyone in the software and legal world, did not think APIs were copyrightable). They are not even bothering to argue that they had a license.
Incidentally this is similar to how Oracle did not comply with that SDKs Apache 2 license... oddly they are trying to argue that they had a license despite that.
https://en.wikipedia.org/wiki/OpenJDK (OpenJDK release date)
https://en.wikipedia.org/wiki/Google_v._Oracle_America (The two other dates)
I don't personally think an API should be subject to copyright, I think copyright and patents in things like this is counter productive. But the law here isn't in alignment with my personal views, and Oracle's defence does not vest in some supposed consistency across things they use and things they own: if they can prove they own a thing which is bound by IPR/Copyright rules, then thats the law.
'Do not steal my property' is not rebutted by 'but you stole somebody else's unrelated property in another situation' -If that was the case, an awful lot of law has to be re-written, its pretty much tit-for-tat law at that point.
Another take would be "even criminals have civil rights"
In what way do I mis-represent the situation? Oracle is suing Google for copying an API. People are saying "but Oracle copied the Amazon API" -I'm trying to say "yea.. nah: not relevant"
https://docs.oracle.com/cd/E15051_01/core.1111/e13777/prodli...
Nope. special Sun licence, reserved. They used Apache licence for other stuff.
Hypocracy does not negate legal rights (unfortunately)
What Sun did, does not alter what Oracle Claim. It might alter what binds them. I'm not a lawyer. If I was a lawyer, I would still be saying (sorry to be so repetitive) that the underlying story here, about the Apache API licence by Amazon, has virtually no material applicability to their case in law, even if it appears inconsistent.
The nearest I've seen anyone show me why its relevant, is the above comment about how Oracle seek an extension of their API rights, which would be in conflict with what the Apache licence inward for Amazon grants them. I absolutely concede that might play in their case, against what they seek. I just don't think its the giant "Ah-HA!" moment. Because. .. Hypocracy is not actually illegal per se.
I may demand you don't kill me, while (stupidly) denying I commit an offence killing you because they aren't the same. I may be held guilty of killing you but it wouldn't therefore make you innocent of trying to kill me, because the one does not negate the other.
Oracle may indeed wind up in breach of the Apache licence over the Amazon S3. Or.. they might cross licence it from Amazon in return for a RAND licence over their claims for the Java API, or hold-harmless or some payment in cash or kind.. Anything is possible. If Bezos hated Google enough, he might even come in on Oracles side.
Did Oracle infringe Amazon's copyright here? Ars Technica contacted Oracle to ask them if they had a license to copy Amazon's S3 API. An Oracle spokeswoman said that the S3 API was licensed under an Apache 2.0 license. She pointed us to the Amazon SDK for Java, which does indeed come with an Apache 2.0 license.
I only brought up the Apache licence, because the article we are talking about, brings up the Apache licence.
Low level protocols like HTTP are arguably different for different roles, which is where I think this confusion comes from. For example, a socket server has to listen, while a socket client has to connect, and those are different processes.
But an application's interface is describing a set of, broadly speaking, function call signatures.
Specifically, it's describing the names of the function calls, the parameter types and return types. (And it describes those types as well.)
Those must be identical for the API to work. There's no difference between the "client API" and "server API" by the definition of API, clients and servers are different in role and implementation.