20
votes
AWS announces they will create and maintain an Apache-licensed fork of Elasticsearch and Kibana
Link information
This data is scraped automatically and may be incorrect.
- Title
- Stepping up for a truly open source Elasticsearch | Amazon Web Services
- Published
- Jan 21 2021
- Word count
- 992 words
Latest in an ongoing back-and-forth about license changes to Elasticsearch and Kibana.
The announcement from Elastic that kicked it all off: Upcoming license changes to Elasticsearch and Kibana
Followed by License Change Clarification, an FAQ, and a much more confrontational tone in Amazon: NOT OK - why we had to change Elastic licensing
Previous discussion on Tildes: Elasticsearch and Kibana are now business risks
As much as I hate the bastardization of open-source licenses, I kinda get it.
Suppose you're trying to be a good player in the market, and offering SaaS to pay the bills. But then comes a giant mega-corp doing their own SaaS with your program, using your trademarked name without permission, and not giving you a cut...yea I can see why the modifications are happening.
IMO a great license would be "GPL, but if you sell the software itself you also must give us a cut."
How much would you split them, though? I think Amazon should be broken up, but among its horizontal slices. So AWS would be its own entity. But that solves nothing for elasticsearch - it's not like it's Amazon shopping that's causing the issues. If anything it's AWS propping up much of the rest of Amazon's empire building since its such a huge margin business.
edit: after more mulling on it I don't think the issue really is that Amazon is too big. I think it's just hard to build a for-profit business on FOSS software, GPL or MIT. What elasticsearch as a company really wants is more of owning a patent - other companies can run it, but they need some royalties. The same thing happened to Docker. Other companies just did a better job at running their OSS product, fully by the books, and it wasn't just one company, it was just every webhost.
Not that companies that focus on managing FOSS software shouldn't exist, but probably a foundation-style non-profit that seeks contributions from companies that use it and is much leaner is more sustainable if you're going to make the core of your product open source.
I don't think AWS is actually that much of a monopoly, though. They're the single biggest player, but at the same time, that's only 32% - Azure is hot on their heels at around 20%, the "field" is still the plurality at 37%. For my solo projects I've never considered AWS and never had to - I use DO and Linode, both smaller minority webhosts which are thriving. If you don't want to use AWS, you have no lack of perfectly good - might I even say, better for most people - alternatives.
But even if you did split it into AWS of the West and AWS of the East, I don't think Elasticsearch is in any better position. It's the same issue Docker has.
They don't have a real competitive edge in the for-profit scene. Their hypothetical killer edge is that as the maintainers and primary developers of elasticsearch, they are the experts at using and hosting it. But the reality is that it's not hard for a team of talented engineers and DevOps to be just as good at that product as the creators.
Like, even if everyone behaves as they should in the GPL sense, where everyone uses and develops the FOSS projects, open sources and upstreams their work back to elasticsearch - that model still involves elasticsearch's paid product being outcompeted.
And I don't think any amount of antitrust will change that.
But as long as elasticsearch is any way OSS, in the GPL sense, I don't see how you can actually prevent that kind of verticality. That is the point of OSS, after all. And I mean elasticsearch basically reached the same conclusion - they switched to a license that is not recognized as an open source license by the OSI.
I think one way to prevent that kind of vertical integration is to separate hosting providers from software providers. Basically forcing a wall between Infrastructure/Platform as a service and software as a service. You can do one, or the other, but not both.
That would lower the barrier to entry for both hardware AND software providers. SaaS providers could ride on PaaS platforms, and PaaS platforms could spin up without also needing to provide software support.
Google/Amazon/Microsoft could still own their hardware, but they would have to choose between hosting only software (Office, Databases, etc), or renting out hardware (containers, VMs, etc).
There is precedent for this kind of thing. Movie production companies were prohibited from operating their own theaters, because they could leverage their copyright monopoly to push out competitors, and eventually end up owning the entire industry.
Other industries this need applied to:
This is it. It's the classic tragedy of the commons: everyone takes the free resource and profits from it, until the free resource is dried up and everybody loses.
Why is Amazing bad here? You could argue ES used free workload to make commercial product, people contributed believing it'll be OSS but suddenly it's closed because ES don't want other companies to host it? Should other free software projects follow the suit? Amazon wants to keep it open source, anyway i don't remember any such fuss when MariaDB was forked out of MySQL, it was actually in reverse
Calling for Amazon to be broken up is weirdly similar to screaming Twitter is censoring
You'll see from my other posts here that I'm kind of in the middle of the road on this issue, but to answer your question directly: Amazon are "bad" for stripping a revenue stream away from the maintainers of a very useful OSS package, in a way that obeys the letter but arguably not the spirit of the license.
It's not that elastic are taking people's contributions for free, it's actually the opposite: elastic are funding the vast majority of the development and publishing it openly for the rest of us, on the assumption that they can make that money back on their hosting services. If other big companies jump in at the hosting stage, the funding dries up and the OSS development dies unless the volunteer community can pick up the slack. Even if the community can save this one, it puts other companies off open sourcing their work in the future.
I don't like that idea because it implies that other people's commercial usage and contributions are less worthy than the original author and I can't agree with that.
I personally think Elasticsearch should've gone with AGPL and so Amazon and other companies would need to publish their own changes and everyone could benefit from it since it's likely these companies would want them upstream at that point.
To me, it seems like both Amazon and Elasticsearch want to keep their proprietary forks alive for their own benefit.
How important are Amazon's code changes to the actual issue, though? I would guess that the work they've done on the software is pretty minimal - the issue is that they're providing expert configuration and cheap hosting at scale, which elastic's own cloud product can't compete with.
I see it as a question of business viability, not code contribution. Elastic funded the development, Amazon are profiting from it at the expense of the original developers. Making Amazon either pay up or stop using is the explicit goal.
There's a very healthy argument to be made that this kind of obstruction isn't in the spirit of permissive licensing, and of OSS in general. There's an equally healthy argument to be made that if the companies behind products like elasticsearch can't fund their ongoing open source development, none of us will have access to the benefits. I don't have a neat answer, but I can say with reasonable certainty that it isn't a clear cut question.
Maybe. But Elastisearch is only the most recent of several doing the same manuver.
Amazon's relation to open source is mostly a one-way street. Google and even Microsoft contribute back more than Amazon. And the smaller operations just can't compete.
Well the AGPL makes contribution the only way if you want to make modifications because it considers usage over the network as redistribution.
Also, you're right, Microsoft and Google have definitely dipped more (although I personally think their contributions are usually for their own benefit) but Amazon does contribute to certain projects as far I can tell. They contribute to OpenStreetMap for example. I guess that fits into "mostly" though.
That just sounds like reinventing patents
A better approach would be to make the software a complement of another product that isn't just hosting.
Customization, training, consultation, migration, consultation, etc. can all help pay the bills.
Let me stop you right there, so I can't name software I host just because I'm bigger than said software developer?
Yes, that's what a trademark is for. Specifically about selling a directly competing product, not necessarily the size of company.
Elasticsearch sells a product called Elasticsearch Cloud, directly on Amazon, Microsoft, and Google cloud platforms. Amazon sells a competing product called Amazon Elasticsearch service.
If I make ibuprofen pills, I can't call it "Vord Advil". CentOS is named as such because they can't use RedHat trademarks.
Amazon even does this for other products. They don't sell MySQL or PostgreSQL...they sell Amazon Aurora, which is MySQL or PostgreSQL compatible.
I don't think that comparison is quite accurate, since elasticsearch is the name of the software, and a hosted instance of that software is what Amazon are selling. The company that makes elasticsearch is just called elastic. In this analogy, Amazon aren't selling generic ibuprofen, they're literally reselling Advil - you can buy the Advil from GSK, the manufacturer, or from Amazon, a reseller.
Aurora is subtly different, it's an Amazon fork that retains compatibility with the original. RDS (relational database service) is the one that uses the core version of whatever DB engine, and the naming convention there is "RDS for PostgreSQL", etc.
But Amazon isn't selling Elasticsearch. They are selling an Elasticsearch service (leveraging the Elasticsearch software) in direct competition with Elastic's Elasticsearch service. Even if it was, trademark for both the name and logo of software is legally protected if it was registered.
Amazon is free to make a product based on Elasticsearch. But they can't brand and advertise it as such, especially without the company's permission.
The fact they've been allowed to get away with it is a bastardization of the legal system.
That's a fair distinction. Perhaps a better analogy would be an independent mechanic advertising that they specialise in Ford trucks?
Surely, in that hypothetical, they would be allowed? Retailers can use the trademarked names of the products they are selling in order to describe them.
My understanding is that they are allowed - it's an informational use of the term: they are literally running an elasticsearch service, and they describe it as such in the name.
I'm struggling a little to understand what you would consider acceptable. Is it a question of prominence? "Amazon Search Service, powered by Elasticsearch", for example?
If you suggest they can't use the term at all, then how would they inform customers of what they're offering?
They typically have permission from the company that created those products. And they typically buy the product from the creator or give them a cut of the profits, not take over manufacturing and cut them out of the equation.
But are they required to? I'm trying to separate out "acting in a shitty way" from "actually breaking the law" - it's a very different matter depending which Amazon are getting away with. If you define the law in such a way that a company can't even resell a product by name without direct permission from the manufacturer, I guarantee that the behemoths on the manufacturing side will use it to penalise small retailers.
This is where the analogy gets strained. I've already said a few times in this thread that I think Amazon's behaviour is likely to directly cause elastic's funding to dry up, so it's not like I'm defending them, but the previous license did pretty clearly allow anyone who chose to to host their own version.
I'd be tentatively in favour of a license that forced hosted service providers (as distinct from self hosted users) into some kind of revenue share deal here, but that's quite separate from the trademark conversation.
Can someone, who is a bit more understanding of free software licenses, explain if AGPL would make Amazon return code to Elasticsearch or other such companies? Is this a solved problem, apart from the fact that they use a non-copyleft license?
If my assumption is correct, why do they (these companies like Elasticsearch, Mongodb) feel that this wouldn't be a sane solution?