IANAL, but I believe in this case a maintainer can just fork MongoDB from right before they changed the license, then new contributors can just contribute to that forked repo, creating an off brand “Non-goDB” that will get a lot of the updates that people want. This would fracture the development of the project and would be bad for everyone, but at least the open source community wouldn’t completely lose the project.

IANAL either but just to be pedantic, in theory I don't see what stops them from no longer distributing old versions under AGPLv3. Having distributed it at one point under GPL doesn't force them to continue doing so forever.

Of course that would be pointless as anyone who forked before the license change could continue re-distributing under AGPLv3.

If you contribute to a project, you have no entitlement, moral or otherwise, to _future_ contributions to that project that others choose to make. Future contributions may be made under different terms, as you point out. The project might stop development, get forked privately, and you may never see active future developments distributed again.

This is self evident from the licences themselves but also clearly reasonable if you consider the proportions of contributions made. Why should a contributor get rights to all future work on a project made primarily by others solely by making a contribution?

> ...and the open source contributors walk away with nothing?

Nope. As others have pointed out, they walk away with the Free Software licensed version of the code to which they contributed, together with their contributions. This code base does exactly what it did at the time of their contributions.

But you can still offer it as a service, just as long as you opensource your infrastructure. The only thing it prevents you is profiting from lock-in of your services without sharing those profits with MongoDB. Very much in the spirit of open source.

Yeah, but how is that really supposed to work? I can see why MongoDB would want to do this, but how on earth is this really supposed to be implemented? Where do you draw the line? What I suspect is that you'll end up with a bunch of shell scripts for creating MongoDB instances...

MongoDB bets that anyone who wants to run mongo-as-a-service will ask them for a business license (instead of open sourcing their whole cloud management stuff), and they will work out a deal.

>> shell scripts for creating MongoDB instances...

_minified_ shell scripts

> sharing those profits with MongoDB. Very much in the spirit of open source.

I'm not sure how these ideas tie together.

I had the feeling that AWS and Azure are pushing their own NoSQL solutions and don't care much about MongoDB.

Azure offers Cosmos DB advertised as a drop-in for MongoDB. I'm guessing Cosmos is Mongo in disguise, though I don't know if it really is, and/or whether MS has made a deal with Mongo or just think they can use the AGPL version.

For the record: I'm not a big fan of Mongo (the DB) but I think MongoDB, Inc. raises a valid point wrt. developers doing all the hard work including community building and a million other things, while "cloud providers" get all the money. This isn't sustainable, and we need a license which more clearly says "if you make money with it, you need to give us some", rather than using the bare AGPL license without further qualifications in the hope AGPL's "freedom" aspirations have the indirect effect of forcing commercial users and/or resellers to pay.

Did the MongoDB pay for their Linux distros? GCC? Git? How much are they paying to the FOSS projects they used to build their software for sale?

They probably report issues back and so on.

How much AWS shares its profits with GCC, qemu, etc? Not at all besides the bare minimum that benefits them.

(And don't misunderstand me, I don't see Mongo as some kind of underdog here, but making big money from software is hard, even if - or exactly because - valuations - and funding - are in the sky.

And AWS is fundamentally a different kind of business than what open source usually is. It has a compounded networks effects / economies of scale effects. And on top of that it has a closed source management layer, but that is largely irrelevant, except for Mongo in this case.)

Fair point, but if you strip any money from product development, this implies you don't have a budget for further development and maintenance, except if you make it support-intensive. And for what gain? That very few cloud providers make more money and enslave customers into their walled garden? Cloud providers don't even need to provide QoS or support - they can just shrug and say "we're running an OS project as-is". If no money can come your way, then development and support is simply not sustainable.

The projects you mention: Linux+git development has good financial standing from companies who could be seen as going (or having gone) aggressively against commercial Unix from a business PoV. gcc thrived on freedom enthusiasm but has seen many, many patches from commercial vendors wanting their OS, CPU or whatever supported.

Intent matters.

Linux, gcc, git OSS projects are less concerned about capturing value from the economic activities enabled by them, which is why they were all set up as not-for-profit enterprises; the expectation is that donations will help cover their operating costs.

MongoDB on the other hand is a for-profit enterprise that took VC funding in the hope that they'd be able to cover their operating costs in the long term without donations.

The only way to do this is to capture some of the value offered by the NOSQL paradigm, but it appears that the cloud providers are beating them at that game, which is why they needed to change tactics, to guarantee their survival as a self-sustaining entity.

That's a great perspective that I had not thought about before.

Fair point, actually.

CosmosDB is Microsoft's proprietary solution. According to [1] started already in 2010 to support their own online services.

[1] https://azure.microsoft.com/en-us/blog/azure-cosmos-db-micro...

I wish it was Mongo under the hood. Our migration to Azure almost crashed and burned because, aside from the extortionate pricing model for Cosmos, their "Mongo API" is noticeably incomplete (in our case, they didn't support the array positional operator, and still don't [0]). Nowhere do they have an API compatibility checklist or anything articulating less than complete support for all features--the positional operator has been in Mongo since version 2.4, five years ago. (version 4 is what's current now).

IMO, Azure remains a garbage fire of half-finished features in almost every offering. We replaced Cosmos with a Bitnami-authored canned VMs pre-configured for a simple Mongo cluster (which make up a significant portion of "features" they offer), which failed to restart after the Meltdown cloud reboot because the Bitnami management software corrupted the disks. Now we just run our own VMs with our own Mongo instances and curse the day we were obligated to move to Azure.

[0] https://feedback.azure.com/forums/263030-azure-cosmos-db/sug...

> I think MongoDB, Inc. raises a valid point wrt. developers doing all the hard work including community building and a million other things, while "cloud providers" get all the money.

I think MongoDB is free loading off the the community and developers. The developers don't get to use the software to make a profit without kicking some money upstream to mongo either. Is Mongo going to start offering money for bug reports and patches or do they get to freeload off of their community?

I think you're overestimating outside contributions to these types of projects. In my experience, they are typically minuscule, and mutually beneficial relationships when they do occur.

A lot of AWS's managed services are just off the shelf open source projects on the backend.

Amazon can afford to fork the project and poach top talent from your company. I think these moves are extremely short-sighted.

Changing license models only causes instability in the overall ecosystem.

> A lot of AWS's managed services are just off the shelf open source projects on the backend.

Amazon employee here. Amazon offers two different types of service. For a service like Amazon RDS then yes it is the standard open source project hosted and managed by AWS. The value add is that you don't have to administrate and back it up, etc.

For something like Amazon Aurora its a different story. Aurora is API compatible with MySQL and Postgres but it is designed internally to work quite differently, really designed for the cloud first. And the result is its up to 5x faster than the standard open source software and just better than the open source version in many ways.

Both as a former customer of AWS and now as an engineer at AWS I have always preferred the from scratch implementations by AWS over the hosted open source versions. I'd rather not use a forked version of open source. I'd rather use something built by AWS engineers from the ground up

> For something like Amazon Aurora its a different story. Aurora is API compatible with MySQL and Postgres but it is designed internally to work quite differently, really designed for the cloud first. And the result is its up to 5x faster than the standard open source software and just better than the open source version in many ways.

It's also noticably slower than plain PG in other cases.

As the cloud often is.

Why doesn't Amazon just pay mongodb? If you you hired some employees from mongo you would have to pay money.

Predictability: who's to say what MongoDB will decide in the future e.g. ask for more money and/or charge per vCPU + memory. Employing developers directly, on the other hand: here's an offer and vesting schedule - they know exactly how much they will spend for 3+ years, and get to control the roadmap.

Because Amazon is all about squeezing every penny out of their supply chain.

Load balancers, DNS, DB, Virtualization ... a lot of those services on AWS are based on opensource software.

I heard that AWS offered mongo XX million dollars per year to have Mongo included as a backend for RedShift. They haven't touched it yet, because of its license.

"Imagine if all projects did the same... Say Docker?...."

Wait just one second here. Is docker still open source or is Moby the new open source docker?

https://blog.docker.com/2017/04/introducing-the-moby-project...

My understanding is that recent versions of docker is indeed not open source but moby is? Is that correct according to the OSI?

wouldn't the older versions of mongodb still be under the old license? (IANAL)

Yup, article also states this.