Hacker News Reader

Comment by jeswin | Hacker News Reader

Elastic Search launched more than a decade back, when the cloud was used in an entirely different manner. The team had no way to foresee the shift to the managed services model; of AWS taking popular Open Source products, repackaging it and undercutting the original authors.

The decent thing to do (for all managed Open Source products that you didn't create yourself), would have been to offer a cut to the original team. Instead, AWS takes all their customers, keeps 100% of the revenue, and passes on nothing.

Terrible.

View on HN

> The team had no way to foresee the shift to the managed services model; of AWS taking popular Open Source products, repackaging it and undercutting the original authors.

The Affero General Public License (AGPL) was published in 2002. https://en.wikipedia.org/wiki/Affero_General_Public_License

Plenty of people saw this coming. It was completely predictable. The reasons why licenses like the AGPL haven't yet caught on are complex. But I don't think anyone can credibly claim that they were blind-sided. Rather, I think many people and companies simply feared the day of reckoning, discounted the threat, and did as most everybody else was doing.

> The reasons why licenses like the AGPL haven't yet caught on are complex.

They are not complex. Permissive licenses encourage adoption but hinder monetization; restrictive licenses do the opposite. Pick your poison.

That was the argument against the GPL, and yet Linux beat BSD. The tension exists, but it's not such a simple trade-off, and the dynamic is highly context dependent. The complexity is in identifying and achieving the best business model.

Like with the GPL 20 years ago, the intuition today seems to be that licenses like the AGPL are simply anathema to adoption (because "viral", etc) and therefore market capture, and if you can't capture the market you can't monetize your investment. It's something of a self-fulfilling prophecy, but eventually more people (authors, users, etc) will begin to explore the space commercially and, hopefully, disprove the conventional wisdom.

> The team had no way to foresee the shift to the managed services model; of AWS taking popular Open Source products, repackaging it and undercutting the original authors.

First release of elastic was Feb 2010. AWS launched in March, 2006. The folks at 10gen used the AGPL for MongoDB since Feb, 2009. Yeah nobody could've seen this coming.

Not sure Mongo is the right example to use here, given the've gone down exactly the same "non OSI approved license" change themselves...

In fact, Elastic new license was created by Mongo.

It wouldn't surprise me if that license in a few years don't get nicknamed "Mongo's license"

AWS didn't undercut anyone. They offered their customers exactly what they wanted. Many of these software vendors have delayed offering their own cloud-hosted versions, often years after AWS, even when these vendors should be better equipped at running them.

Elastic is also infamous for data loss and still has consensus problems so maybe proper priorities are the real problem.

There is no meaningful way for an independent vendor to compete with a Cloud Provider offering the same service. It gets bundled in the initial sales pitch, promoted in their documentation and at events, and gets heavily discounted.

It simply becomes the default - and (if on AWS) nobody will even think of Elastic after a couple of years.

You compete by not competing. Make your software proprietary and sell it to the cloud provider for hefty license fees.

And yet many companies are competing just fine. Elastic's biggest growth comes from cloud services, which matches every other database vendor from MongoDB to Redis. [1].

AWS just proved that the market wants to buy solutions, not software. There are tons of customers and environments where a dedicated vendor can offer something far better than the lowest-common-denominator that AWS provides.

If you want another success story, just look at how Snowflake built a better Redshift on AWS (quite literally).

1. Start at page 23 of their earnings report: https://s2.q4cdn.com/265747582/files/doc_financials/2021/q2/...

The big three may have some advantages, but there are numerous smaller competitors.

The biggest advantage the big three have is that some people don’t even bother looking for alternatives.

> The decent thing to do (for all managed Open Source products that you didn't create yourself), would have been to offer a cut to the original team.

Who are the members of the original team? Are we not going to give a cut to all the other contributors to the project throughout the years? How much should each contributor get? Are we measuring by lines of code touched?

If Linus had hindered AWS' use of Linux early on, would that have been better for Linux? What would "the cloud" look like?

They don't seem to have any revenue sharing scheme with Linus, yet all the Linux companies seem to be doing pretty well for themselves.

Linux is not project developed as a product mostly by a single company. Substituting Linux into this situation doesn’t really provide any insight.

Actually, your argument that there’s any relevance to Elastic being the primary developer lacks insight. It changes nothing about the licensing situation nor about the ethics. You’re also not factoring in the fact that the nonrestrictive license is why Elasticsearch became the standard.

You also haven’t resolved the logical contradiction that Elastic is equally “guilty” of profiting off of Lucene.

I honestly don’t think a nuanced understanding of these licenses and open source business models can lead to any conclusion except this one: Amazon is not in the wrong here, Elastic is being entitled, misleading and duplicitous. Elastic is free to use Lucene as its kernel and monetize it, and Amazon is free to offer a managed Elasticsearch service.

If you write Apache 2.0 software, you have no right to bitch about competition. That’s the short of it. And remember - Elastic is doing fine. They’re worth $15B. The founder has hundreds of millions of dollars now, and rightly so.

Right. So what would Linux look like if it was developed by a single company?

One of the reasons to choose open source is to avoid lock in by having multiple suppliers, and at least in theory have some healthy competition between them.

> "We proposed to give back jointly at a significant level and were turned down." Adrian Cockcroft (VP AWS)

https://twitter.com/adrianco/status/1105178074499375106?s=20

We have no idea what “significant level” means here.

~1k lines (per enumerated prs in blog post) against millions of lines, aka less than a fraction of a percent.. plus ignoring code review, issue triaged etc. insignificant level might be more apt.

Amazon is literally under no obligation to contribute any code back. Yet they do because most people prefer to merge upstream if they can, and Amazon is no exception.

Frankly Elastic has been very hostile at accepting PRs if they seem like they’d compete with Elastic’s proprietary offerings. Glaring conflict of interest. So I’m glad Elastic forced a full fork.

BTW, in case you didn’t realize this license change will not prevent Amazon from running their own fork. So even from a greedy business perspective Elastic made a horrible decision, even ignoring the ethics of lying to the community as they have done repeatedly and without showing any remorse.

edit: Fixed typo where I wrote "Elastic" instead of "Amazon" when talking about Amazon being no exception to preferring to merge upstream where possible

There's a good quote on that thread which captures what AWS is doing with products like ElasticSearch:

"My argument is that AWS is eating seed grain, on its supply side, and doesn't know it."

AWS launched RDS with MySQL 4 months before the first version of ElasticSearch. Yea in that case it was Oracle, but I think it was clear what the direction was. And it was definitely clear in 2014 with Aurora and Redshift. They could have changed their license in 2015, but they didn’t. Why?

If it was around 2010 MySQL was not yet Oracle. (Maybe not even yet Sun?)

> Elastic Search launched more than a decade back, when the cloud was used in an entirely different manner. The team had no way to foresee the shift to the managed services model;

This claim is highly unrealistic. Offering managed services has been the hallmark of the cloud from day 1. The cloud is nothing more than managed solutions.

It's highly disingenuous to claim that it was not possible to expect a business dedicated to offer managed services to offer a managed service.

I generally agree, but isn’t the ability to do unforeseen things one of the powers and major wins of being a user of open source ecosystem?

I don’t disagree about, what I would say charitable, instead of decent thing to do is.

The issue is, I think, that acting in such a manner is supposed to cost you in community good will. The potential for bad PR can force you to act in a more ethical manner, but there's a big separation between the people with the checkbooks and the folks who are aware of the issues here. The courts are there for extraordinary circumstances but it's generally better for all parties to settle a dispute without litigation and AWS could just shovel some money in the direction of elastisearch.

Reply on news.ycombinator.com