DEV Community: Nicolas Gimenez

Which business model to choose for your next FLOSS project?

Nicolas Gimenez — Fri, 14 Feb 2020 18:04:56 +0000

How do we financially sustain the developement of a FLOSS? What are the advantages and drawbacks of each approachs, based on real-life examples?

If you are unconfortable with the notions of open-source, free software and software licenses, read the first part of this series!
For my humble thoughts on the ethical stance of both movements, check out the second part of this series.

The choice of software licenses is key to a successful business model. As retroactively taking back rights given by a license is not possible, it must be thought ahead.
A few technically successful projects have been abandonned because the maintainers couldn't find time to work on the project anymore.
Sometimes, even companies which engaged into raising money from venture capitalists did not succeed financially despite the widespread usage of their product. A good example of that is Docker.

There are plenty of initiatives these days to pay the open-source developers, namely GitHub Sponsoring system, Patreon pages, RedHat Tidelift.

Selling the software itself

Free standalone Softwares may be sold the same way as proprietary softwares. As said earlier, nothing forbids it. The most famous example is copies of GNU Emacs, that was sold by Richard Stallman back in the time.

But GNU Emacs was distributed at a time when barely 10% of the global population had access to a very slow internet. Today everyone can download it for free. Even if you'd lock Emacs and release it as a free trial with a paid option, someone could study the source code and release another version without the lock for free on the Internet. All this is legal and encouraged by the FSF.

That being said, non-technical end-users do not care about software licenses. They do not know anything about it. If you tell them to pay to use the software, they will just do it. But they might feel betrayed if you did not warn them that there were a way around payment.

Generally speaking, selling the software itself does not work well if the software is 100% open-source. It has better result when at least a tiny part of your software is proprietary.
A counter-example could be the commercial "premium" WordPress plugins which are free softwares (licensed under the GNU GPL).

Support, training and donations

Most people distribute FLOSS for free, and make end-users pay for advanced support, specific training or simply ask for generous donations.

If the software is a standalone as opposed to a software library, the main interest of the license business-wise is to enforce sharing improvements. Therefore, you'd rather use a strong copyleft license - such as the GNU GPL.

Ex of projects adopting this strategy:

the GNU projects
Linux
WordPress
Blender

For a software library, using a permissive license can be interesting because:

it makes it smoother for people to use the library
it allows the software owners to both accept contributions and make proprietary add-ons on top, which may reasure the investors behind the project

Ex of projects adopting this strategy:

Some prefer using a weak copyleft license instead for libraries:

VLC core library which recently moved away from GPL to LGPL
ZeroMQ which uses LGPLv3+static exception and is moving towards MPLv2

It really depends on your goals:

are you pursing the Free Software spirit or the Open-Source ideology? ==> go for copyleft for the former
do you want to enforce sharing of improvements, avoid being captured by a proprietary project ==> go for copyleft
do you prefer making sure everyone can use your software? => go for permissive

It must be noted that if somebody forks your permissive project and make it GPL, you will NOT be able to pull back the improvements to your permissive project using your permissive license. You will need to change your license to GPL.
That's one of the reason why permissive license advocates often dislike copyleft users. They consider that GPL-like licenses are viral and polluting.

It's a risk and this reason alone is enough to advice any one of you wanting to engage into making a significant open-source library with a permissive license to choose a weak copyleft license like the MPLv2 instead, especially if you are a small startup or an individual.
Pieter Hintjens talked about it better than me on his blog.

Martin Sústrik, ZeroMQ co-founder that forked the LGPL project to create Crossroad I/O, famously disagreed with the use of a copyleft license in ZeroMQ. He explains why he is a permissive license supporter on his blog.

As a tolerant Free Software advocate, I'd recommend anyone that is forking a permissive library to a copyleft-licensed repository to be a good OSS citizen: if you can, push your changes relevant to the original work to their original repository with their own permissive license. Let's not impose our views on people.

Interestingly, from what I've seen, the two projects of Martin Sústrik, Crossroad I/O and nanomsg, did not (yet) have the same sort of success, and did not gather as much of a community as ZeroMQ did. Maybe the use of the "benevolent dictator" model for handling his community over the innovative de-centralized C4 contribution policy of ZeroMQ can explain it. Or maybe the choice of license? It is hard to say.

At this game of selling support and training, it mostly doesn't work for individual developers and small companies. RedHat is the real winner here. The land page for their Tidelift project talks about it better than anyone. Most companies don't want to bother with a tons of different support contract for tons of different software modules - and as RedHat says: "Open source creators: RedHat got $34 billions of you got $0.". Individual or small group of developers are no lawyers and salesmen, and that's what it seems to take for companies to sign legal and technical support deals.

The amount of money received from donations for small projects are usually only enough to pay the founder/core maintainers of the project. Anyone else contributing, even significantly is not paid at all. The most successful open-source software working with that model is Linux.

This interview with the VIM creator Bram Moolenar is symptomatic. When asked "How can the community ensure that the Vim project succeeds for the foreseeable future?", he answered: "Keep me alive".
Often times, when an open-source software relying on donations to a one-man maintainer becomes that successful, the maintainer is hired by a well-established company to work on the project full-time. Bran Moolenar works at Google.

What happens when the main project maintainers can no longer contribute to the project? It seems quite scary for the users that rely on such library... That's why it is a necessity to create a community of developers - and somehow find a way to get them paid, to some extent like with Linux.

MaterialUI makes it very clear where the money goes and seems to split it to many contributors.
Still, it is difficult to pay everyone a competitive salary. One of the MaterialUI core maintainer, Olivier Tassinari, acnowledges that and as a result is part of the RedHat Tidelift subscription.
In general, FLOSS are a lot more complicated and complex than proprietary softwares, and thus, such developers would be paid quite a lot if it was in the corporate world.

Another problem with relying solely on support (and to some extent, training) is that the more your software is great, the more the documentation is great the less people will need your training/support and therefore it can create a paradoxical vicious loop: the more you get money, the more you can improve your product and the less you will get money.

Open-core

The fundamental idea is: the core of the product is open-source, but more advanced, "entreprise-grade" features is proprietary pay-to-use software.

An example can be JetBrains IDE products and their freemium business model. A free open-source "community edition" licensed under a permissive Apache 2 License and an "Enterprise" proprietary paid version.

Another alternative is distributing a core open-source product together with another software like a GUI to facilitate the usage of the core. Ex: MockLab built on top of WireMock (even though MockLab is also a SaaS - see later).

The FSF considers any of this behaviour harmful to the cause of the Free Software, obviously, even though they do allow it through their LGPL.

It can feel very arbitrary to choose which feature should be part of the "Community" edition and which one should be part of the "Proprietary" Edition.

Building proprietary commercial softwares on top of an open-source project can be a deterrent to gathering a community around the core project. This strategy is usually supported almost entirely by the company that initially founded the project, and often require raising funds from venture capitalists.

A counter-example could be GitLab, an open-core alternative to GitHub that is relatively well supported by the Free Software community. It seems like the Debian community is contributing to the Community Edition of GitLab.

SaaS

SaaS stands out for Software as a Service and is a practice of hosting a Software behind a Web Server for a user for a fee.
The Free Software Foundation does not recommend anyone to use a SaaS because they take away fundamental freedoms from the users.

That being said, it is a common and pretty profitable business model these days - if not the most. It's also a huge convenience for the user and I am not myself against Cloud services per se as long as the Cloud provider is transparent with what it does.
WordPress uses this business model extensively. WordPress.org is the Free Software community resources for self-hosting. WordPress.com provides a managed-version hosted in their private cloud for a fee.

As WordPress is GPL, hosting a modified version of WordPress behind a server does not trigger the copyleft effect as the software is used in a private server and not distributed in the traditionnal fashion.

That enables many other cloud providers to provide hosted version without abiding to the copyleft. It created a healthy competition for the users that can choose the best and the cheapest option. It also encouraged external developers to write plugins for WordPress.

For WordPress, this model has been rather successful, as they also sell training, support, ask for donations and sell "premium" plugins. The word "they" here, refers broadly to the WordPress community as a whole:

the WordPress core developers
the independent third-party plugin developers
the cloud providers
the IT teachers
the companies and their developers helping end-users making their website

As a result, the WordPress project founders managed to create a large pool of contributors to enhance their project, especially via their plugin system. These very same people market the project for them, which indeed attract more users and more contributors. This strategy happens to scale very well.

This business and licensing model is the core reason why WordPress succeeded as the most used CMS rather than its early competitors. Not the initial technical value of the product.

I think this strategy works for WordPress because it is a generic Web Framework with an extremely large pool of potential end-users. It wouldn't be as successful if its purpose was more niche like if WordPress was a messaging library addressed specifically to certain developers.

For example, it seems that some Open-Source companies relying mainly on Managed version of their library on their Cloud for getting income have been financially impacted by certain big Cloud Infrastructure Vendors. For example, AWS was proposing the same hosted version as MongoDB without significantly giving back to the project itself.

To avoid this problem, Neo4j and MongoDB initially chose the AGPL license for their core librairies.

However, AWS does not modify the open-source software licensed under AGPL that they provide through their Cloud. They solely distribute it unmodified behind a network. Therefore, they do not need to make available their own code.

This led these open-source vendors to use source-available license for their core products or their additional features, like Redis with the Common Clause and later the Redis Source-Available License, and MongoDB with their Server Side Public License.

Neo4j decided to engage into making proprietary closed-source product on top of their GPLv3 core product.

Another interesting example is ElasticSearch which decided to mix open-source and proprietary source-available code into the same repository.

Dual licensing

Dual/Multi Licensing consists in:

releasing a software library under a strong copyleft license such as GPL or AGPL
selling a "Proprietary License" as well to companies that make nonfree softwares

The project founders shall make people sign a Contributor Software Agreement to all the external contributors to the software for them to give away their copyright to the project founders.
That's the only way they can then publish the external contributions under a different license (the Proprietary License).

Weirdly, sometimes the companies wishing to engage into making nonfree softwares do NOT really need to buy a proprietary license, as some of the libraries/drivers to connect to the core library that's under GPL/AGPL are licensed under a permissive or weak copyleft license.

However, the official vendor website that is selling the "Proprietary License" sometimes voluntarily confuses people by calling such license "Commercial License" instead. The use of the term "Commercial" is confusing because it indicates that you need to buy a license whenever you use the library in a commercial product. That's actually not the effect of copyleft.

Many people seem to think that using GPL means that you got to make your code publicly available. Using GPL means that you have to deliver your source code to the user you are distributing your software to - not to the public. It contributes to confuse people about the intent of the alternative paid license.

Three of the most famous companies engaging into dual licensing are The Qt Company, MySQL and MongoDB.

MongoDB says upfront that their drivers are licensed under the Apache License v2 so that you do not need to license your own code under any specific terms when using MongoDB, but still - companies feel the need to buy a "Commercial" License. I am not sure why, maybe because they need to buy a liability.
MongoDB AGPL/SSPL library was chosen as a protection against big cloud vendors, not to enforce dual licensing.

Qt is less clear. Most of their library are using LGPL so using them is no problem, but they seem to confuse people a lot.
Qt used to be full GPL so there were more incentive to buy a "Proprietary License". Weak copyleft is not really enough to push people to buy a license.

MySQL dual licensing business model worked pretty well until the Cloud became a thing.
This talk from MySQL's founder is very insightful on why they used dual licensing. They make it clear that solely relying on selling services around the product is difficult financially. Dual licensing seems to potentially bring good revenue stream.

Dual licensing is sometimes referred by its detractor as "having your cake and eating it too". For some, it creates "a weird asymmetry and might be a strong deterrent to any community contribution". Pieter Hintjens also despised such practice.

Some say that dual licensing are unfriendly to startups, who cannot afford the cost of the license. Employees in big corporations who may want to use the library do not have credit card privileges too. So at the end it can be difficult to simply get used. Paradoxically, you need that widespread usage to create the public buzz that would convince people that using your tool is worth the money.

However, this StackExchange answer from the founder of iText explains well why dual licensing can be a profitable way to support the developers of open-source libraries.

Conclusion

Business models around FLOSS are really complex. Selling the software itself, Open-core, SaaS and Dual Licensing are just a few of them. I am no expert and I do not pretend being exhaustive here, but I hope this episode will have been useful to you.

Towards a pragmatic approach to the spirit of Free Software

Nicolas Gimenez — Fri, 14 Feb 2020 17:18:54 +0000

What are the important values Free Software and Open-Source carry, and how do they complement each other? I'll give you my two cents, in all humility.

If you are not totally comfortable with the notions of Free Software and Open-Source, read the first part of this series.

Why the Free Software philosophy matters

The Freedom for the users to control their softwares is particularly important nowadays with the ongoing debate around data privacy and the ubiquity of devices. Can users simply rely on the ethics of a privately-own for-profit company? Surely not. Especially not today, with the war of information systems among big nations. It is highly possible that spywares are everywhere from your computer to your phone and your IoT device, some of which are probably installed by the infrastructure vendor itself.

For more information, visit Richard Stallman personal website for an opinionated view on the question. Although he might seems a little paranoid at times, one must acknowledge that this man is full of wisdom when it comes to technology ethics.

"Total surveillance: good or evil?" is an interesting summary of the situation. Edward Snowden disclosed that to some extent, the big tech companies are lying to our faces. I'm not necessarily against all forms of surveillance - but I believe that it should be decided conjointly with the whole society instead of making us trust What's App/Skype/whatever-messaging-app to spy on us.

I want to believe that the majority of the tech companies which do not abide to the Free Software ideology still try to be fair and transparent in their use of our data. However, I don't think it is reasonable to simply "trust" these companies with our data.

With the rise of AI and data mining, the Free Software ideology is particularly relevant. If AI remains as the property of privately own for-profit organization, it's likely to be used for Evil at some point. The threat should be taken seriously. Those who say the opposite will have to take responsibility for their actions and related future events.
I do know we are far from creating a real human-like intelligence but simply gathering and processing data without true consent is highly questionable.

Free Software dimishes the risk of malwares by giving control to the users, but the only way to be sure that you don't run a malware is to build and deploy a source code you fully understand on a machine that you fully control, a way which is apparently.... not for everyone.

That is why I am proud to support the cause of Free Software.

Why the Open-Source movement matters

Technically speaking, softwares made in the open are, as opposed to closed-source software:

less likely to be thrown away 1 month later.
more likely to have better performance as well as security, since more pairs of eyes are looking at them. Management around open-source is de-centralized and more technology-focused.

In my humble opinion, the main problem with proprietary softwares is that we may not be able to know what they are actually doing. We cannot verify in the source code. To me, the essence of the Free Software movement is the availability of source code. It is less important if you can modify the source code, reuse it or sell it. For example, if you have access to the Facebook/Slack source code, at least you'd make sure that they are doing nothing wrong with your data, that they are doing only what they are supposed to.

If they don't, at least you'll have the proof and you'll be able to report them and stop using their softwares for good.

A possible flaw of the Free Software movement is that it is not easy to make a great product and design a profitable business model solely with GNU GPL.
Some business models work, but they are either limited or require lots of creativity.
More often than not, the use of proprietary software, either closed-source or source-available, is necessary to generate higher revenue to support and improve the project.
The Open-Source movement acknowledges that.

Besides, Open Source acknowledges that sometimes, proprietary softwares simply do the job better than their Free software counterparts, especially in terms of design and ergnomics - an aspect sometimes overlooked by the Free Software community. Take Microsoft Word and OpenOffice.org Writer as an example. Microsoft Word is really in advance in terms of design and ergonomics to me, no offence intended to the OpenOffice.org development team. Working against a giant like Microsoft is tough, especially when you rely mainly on Apache Foundation donations and cannot afford hiring as many designers.
Big up to the development team of GNU GIMP 2.10 for being the perfect counter-example to that point. Imho, GNU GIMP is better than Adobe Photoshop in every way, including ergonomics.

Open Source values convenience over ethics, and I think, to some extent, it makes sense. Why choosing a lesser Free Software over a solid proprietary product that already exists in the market? I think it is ok only if the source code of the said proprietary software is made source-available, at least to users or third-party security experts. I do not trust closed-source, and I am even considering that closed-source should be illegal. But for most people out there, softwares are just tools. They don't want to loose time installing, configuring and so on. They just want the software to work so they can focus on their business.

In practice I am still using plenty of closed-source sofware and I won't stop anytime soon, simply because they are too wellmade and I can't afford the inconvenience of not using them - at least for now. However, when I am building one software for my own company, there is no way I build anything else other than a Free Software or a software promoting Free Software, if I can find a valid business model to support the project.
If I can't, I will at least make the source code available to my users, - and/or I will ask third-party security experts to verify my source code (like 1Password does).

The Free Software movement tends to ignore business issues and be too political

The Free Software movement is very political. I do not agree with Stallman's view that "owners" should be placed at the same level as anyone else. "Owners" deserves credits and retribution for their work.

Stallman fights for software freedom to the stage of applying it to all media contents, like movies and books. This is a bold position that is in no way compatible with the reality of the market.
In China, film directors are starving to death because they are abused by their producers. Making a movie demands tremendous efforts and a hell lot more money than making a software. Unlike software developers, filmmakers never ask viewers to "support them in deploying their movies." There is no sustainable business model other than selling access to watch the movies, and in no way should a director be left with nothing but debts after releasing the film.

In fact, another chinese director is "begging for survival"(link in chinese language) as he spent 6 years making an animation that was leaked immediately after it was released in cinema.

Netflix did push the production of lots of independent and original contents that would have never have been made and released without the money. And what is Netflix but a sort of modern pay-to-watch cinema? At the end of the day, it is viewers who fund movies and this cannot be different.

Similarly, a writer is entitled to get the benefit of his work, and I don't understand why Free Software advocates like Pieter Hintjens and Richard Stallman are systematically giving away the "source" of their work online for free. What's the point? Most people will just not buy your books and you will just have worked for free. Why volontarilly becoming a slave of modern society?

Back to software.

I believe that ALL softwares should be freed and most should be publicly available as open-source. It's about making the world a better place to live in. Start making software that people genuinely need. It has been said that softwares are the catalysts of all other industries.
At this point you may say that it could apply also to books and movies.
But no. Because software developers can always be paid to make customized solution or to help with the deployment process (until AI replaces them, but that's another issue ;)).
As explained earlier, proprietary softwares can have hidden functionalities, but a book or a movie cannot hide itself.

Stallman said in a conference, "the price is a side issue, I don't care at all!".
I strongly disagree. This is exactly why lot of people don't make Free Softwares.
Stallman then takes the example of GNU Emacs, a Free software he sold in the past, but that was when barely 10% of the global population had access to the internet.
You don't see many Free Software sold nowadays.
You even witness people getting mad that someone dared selling OpenOffice.org on eBay.

While the FSF likes to say that Free Software means Free as in Freedom, not Free as in free beer, in practice, it is really more difficult to make money with it than with proprietary softwares.
In fact, at the very beginning of the movement, the two meanings of "free" were not explicitely distinguished.

I am under the impression that Stallman doesn't take business issues seriously enough, as he insisted later in the same conference that "The price is a minor detail, I care about Freedom".

Free Software makes it hard to enforce payment to buy the software. We will see later why.

Enforcing payment is in no way unethical in my world. Softwares take many smart people a large amount of time. Besides from developers, a great software needs designers, managers, sales team, lawyers, and other experts (doctors if it is for a healthcare software for example).
Free Software advocates often ignore that they need all these other people, which is why many free softwares have lower UX.

If ALL softwares were free, it would for sure change the game, but for the greater good:

developers will not have to work on crappy proprietary code that will be thrown away in 2 months. Concurrent companies will be encouraged to put together their effort in contributing to the same source code. As a result, we save both money and time, and the users will have access to more unified solutions, less systems to interoperate with.
users will have control, they will really know what they use.

Things go well if ALL softwares are free. But if even one single entity in the market has the right to make proprietary softwares, it becomes unfair.
That's why finding the right business model to support the development of your software matters. Finding one that's both ethical and financially sustainable is a huge challenge. That's "a pragmatic approach to the Free Software ideology" is about.

Conclusion

Free Sofware is a political ideology that goes way beyond the realm of technology. The whole point is to give freedom to the users so tha they can make sure they don't run malware/spyware. Open-Source is a practical approach to software development that acknowledges that working in the open makes better products. The Open-Source ideology embraces more easily being combined with proprietary softwares when it appears necessary for the business viability of a venture. Free Software does it only as a long-term strategy towards its ideological goal.
In the next episode, we will talk about the different sort of business models you can adopt to support your Free Libre Open Source project.

Getting rid of the confusion around Free Software, Open-Source and software licenses

Nicolas Gimenez — Fri, 14 Feb 2020 16:32:13 +0000

I've spent a couple of months reading all sorts of articles and news about Free Software, Open-Source, licensing and more importantly the different business model strategies you can adopt to support financially your next amazing Free Libre Open Source Software (FLOSS) project.

I want to summarize in this series of articles the most important things that I've learnt and that I think everyone in the industry should know.

These articles are beginner-friendly. You do not need any background in software licensing, or even in software development to read it. In fact, my wife who is not in technology in any ways, was able to comprehend its content.

I'll start by a few definitions. If you are in tech, while it may appear trivial I encourage you NOT to skip that section, especially if you consider yourself as a junior/mid-level developer, but it also applies to many senior developers. I've read many mischaracterizations and misunderstandings on the very meaning of Free Software, Open Source and the different licenses to know that you might be confused about these notions maybe without realizing it. As a matter of fact, lots of tech/business articles happily make these confusions everyday.

Free Software

Free Software is a term invented by Richard Stallman.
It was invented at a time when almost every piece of software was proprietary. No one was ever sharing source code. The software industry was following the same principle as the other industries: you make a product inside your company, you sell it and you keep your recipe ultra secret. A business model based on selling ultra-expensive perpetual licenses.
Free Software's motto is "Free as in Freedom". Freedom for the user as opposed to the developers.
It all started when Richard got frustrated after his printer couldn't work and he realized he didn't have access to its source code to fix it. And everyone was refusing giving him the source code because they had signed a NDA. You can read the whole story on gnu.org.

What are those freedoms?

First, Freedom Zero is the freedom to run the program for any purpose, any way you like.
Freedom One is the freedom to help yourself by changing the program to suit your needs.

Freedom Two is the freedom to help your neighbor by distributing copies of the program.
And Freedom Three is the freedom to help build your community by publishing an improved version so others can get the benefit of your work.

Free Software DOES NOT mean the software is free as in free beer. Nothing forbids you to sell your software for 1000000 dollars or give it away for free if you wish. It does NOT mean "non-commercial".

It is defined as opposed to so-called proprietary/non-free softwares, that was particularly common back at the time (and still is today...). It has NOTHING to do with closed-source as opposed to open-source softwares.

The Free Software Foundation (FSF) was created to promote the ideology of Free Software and to promote making Free Software.

You can read on Business Insider a recent example of misrepresentation of the Free Software ideology (Free confused with free beer as opposed to Freedom + Stallman wrongly related to open-source): "He [Stallman] pioneered the concept of free and open source software (FOSS), whereby any programmer can create, contribute to, and give away software for free — offering viable alternatives to corporate-owned and created software."

There are hundreds of similar example of mischaracterization since the start of the movement.

The GNU Project and the Linux Kernel

In particular the FSF was raising funds from donations to hire developers to work on making a Free Operating System. It is the GNU Project.

An operating system (OS) is made on hell of a lot of different components, one of which being the kernel.
Linus Torvalds had already created a kernel (a hardcore chunk of software) that was working well and that was also a Free Software: "Linux".
Linux was used in the GNU OS and the GNU OS was finally released under the name "Linux" or "GNU/Linux". There is a controversy on the name of the OS: the FSF encourages people to use "GNU/Linux".

Copyright and software licensing

First of all, as creator of a software, you hold the full copyright over your work. You can distribute it under any license, and you're NOT legally entitled to follow the license you are choosing to distribute your software to, because you hold the copyright of the software. That's something many people are confused about.

You choose a license that says how OTHERS can use your software. Which rights and obligations you grant them when publishing your source code.

As creator of the initial project, by default you do not hold the copyright for the code the external contributors sent to your repository.
As such, you may not have the right to do anything you want with their contributions. It depends on the license of their contribution. If the external contributors pushed their code to your repository on GitHub for instance, then the same license you chose will apply to their contribution.

However, there is a way to make external contributors hand-over their copyrights to you: the Contributor License Agreement (CLA). You can force them to sign this document before accepting their pull-request. There are many types of CLA, it is rather complex so I won't dive into it just yet.

GNU GPL

In order to enforce the 4 Freedoms of Free Software, Stallman created the GNU General Public License (GPL).

At its core, it does the following:

you can do whatever you want when using the software for your own internal use.
when distributing the software to somebody (as a commercial transaction or not), you must give them the source code. You cannot only give them binaries.
then, the entity receiving the program is free to study the source code, modify it, sell it, redistribute it. They can even publish the source code online, but they don't have to and nobody can demand from them to get their modified version. They are free to keep it for them if they want. When redistributing the software, they must give the source code as well.
anyone that make a modification to the GPL software keep his copyright to the portion of the code corresponding to his contribution
if you make modification to the software in any way and use it for your internal use, you can do whatever you want.
if you make modification to the software in any way and distribute it to somebody else, then your modified work must be licensed under the GNU GPL too and as such you must distribute the source code of the modified work to the person/entity you are distributing your modified software to.

The last point is the so-called "copyleft" effect. It means that by using a GNU GPL licensed software to make a modified work, you are somehow impeding your copyright over the modified work in the sense that you are obliged to use the same license GNU GPL when you distribute it. You will keep your copyright over your own work, but it is the result of the contract you accepted when you started using the GPL software to make your own.

The GNU GPL is sometimes refered by its detractors as "restrictive" and "viral" license. Developers tend not to understand why the Free Software guys talk about "freedom" all the while "restricting" the way newly created software is being used. That's because it is about freedom of the users. It is about freeing the software from one particular vendor, as is the case for proprietary softwares. It is also about freeing the software from the developers and transferring this freedom to the users. Not only freeing it from the initial developers but also all the subsequent ones - that is the very reason why copyleft has been created: to make sure any modifications of the software remain free. So that the source code can get its own life.

All the softwares released by the GNU Project are licensed under the GPL.
Linux is also licensed under the GPL.

As you can see, by definition, GPL softwares are incompatible with proprietary softwares ;).

You may wonder how the heck can you make sure that nobody breaches this license, since the code lives its own life out there in the wild.
Well, this is a real issue. The FSF gives legal support to anyone witnessing a GNU GPL infrigement.
Sadly, many companies seem to embrace GPL violations, as with this 3D Printer or many Android-based OS.

Open-Source

Open-Source is a very different notion. It appeared later.

Linux, the kernel of the operating system GNU/Linux, was using the GNU GPLv2 as well. It was also a Free Software (which is why the GNU Project team accepted to use it as a replacement of their own kernel under development). However, its founder, Linus Torvalds, was not part of the GNU Project nor the FSF by itself.

The MIT and UC Berkeley were also working on softwares used in the first entirely open-source operating systems, the (GNU/)Linux OS.
Why open-source? Because X Window, developed by the MIT is under X11 License (~= MIT License) and the networking softwares developed by UC Berkeley is under BSD License which are permissive licenses (see later for a definition). They were not part of the Free Software movement. Otherwise, they would have used the GNU GPL ;).

People working on GPL softwares were not necessarily always aligned with the ideology of the Free Software Foundation and Richard Stallman. When asked by Wonderview Productions in 2001 "What is Linux relationships with the GNU project?", part of Linus answer was "Well, there is relationships to GNU on kind of multiple levels, one is just the philosophical level of thinking that making your source open is a good idea. [...] You can think as Richard Stallman as the great philosopher. You can think of me as the engineer.".
He refers to the FSF philosophy simply as "having the source open" rather than for the Freedom of users.

More recently, Linus opposed to the adoption of the latest version of the GNU GPL (version 3) partly due to the disagreement with Stallman's view on tivoization, a modern way to restrict the change of a copyleft software by locking it in the hardware itself. It is most likely used extensively by all portable devices: that's the reason why VLC remains at LGPLv2.

GPL does not enforce sharing per se, but the fact that anyone receiving a GPL licensed software can release the source code to the public makes it practically impossible for anyone to keep his work secret and make money out of it. That argument was valid only for non-internet based softwares but we will come to that later (GPL loop hole kinda closed by AGPL).
As NDA over GPL work is illegal. Even though the person receiving the GPL software may not want to distribute it to the world, it became common that anyone making improvements to a GPL project was contributing it back to the community, especially with the rise of Version Control System online like GitHub, which made it incredibly easier to do so. Instead of having to distribute the source each time to each customer, you can put the source online with only a link.

People started realizing that Free Software enables online community to gather, to work from all over the world and to create incredible projects. With the Linux community, led by Linus Torvalds who thought "making the source open is a good idea," Open-Source movement was born. The term was created by Eric Raymond. Read "The Cathedral and the Bazar" for more info.

Open-Source is more about the technical side of the Free Software movement as opposed to the ethical and philosophical stance that the FSF insists on. It is really about the convenience of working together on a commonly shared source code. The license used is solely a tool for either enforcing the sharing of improvements or enabling more users to use the source code (permissive non-copyleft licenses - see later). It becomes a business model strategy.

Open-Source movement does not see any harm in combining proprietary and "open-source" softwares, since Open-Source is only a tool to achieve a better product, leveraging free (as in free beer) labour from all over the world. Open Source contributors are often passion-driven, especially back in the days. Open-Source is also a marketing tool used by businesses today to promote their softwares, penetrate more markets and encourage clients to use their products so that they can sell advanced "enterprise" features later. They may or may not use GPL to achieve such purpose. Since then, the use of GPL for a particular package was no longer enough to differentiate a project from the Free Softwares movement. The GPL can be used solely as a tool for business purposes (like with dual licensing - more about that later), to restrict the usage of a key library, or to enforce improvements.

FLOSS is Free(Libre) Open-Source Software and it refers to both Open-Source and Free Software indistinctively.

Source-available

A software is called "Open-Source" only if its source code is distributed to the public under a license validated by the Open Source Initiative. There are many criterias for a license to be accepted as an "Open-Source" license.

A software whose source code is distributed to the public under a license that is NOT validated by the Open Source Initiative is called "source-available software".

An open-source software is automatically source-available but a source-available software is not necessarily open-source.

Permissive vs Copyleft open-source licenses

There are many kind of Open-Source software licenses. We can classify them into copyleft and permissive licenses.

Copyleft licenses

Copyleft is often understood as follows:

the modified work of the copyleft-licensed software has to abide by the same license
when distributing the copyleft-licensed software, the source code of this software must be distributed too

Strong copyleft licenses

The GNU GPL is a strong copyleft license because ANY modified work must also be distributed under the GNU GPL.
The Affero General Public License (AGPL) is the same as GPL except that the event that defines "distribution" is extended to "making your software available to a user through a Network". It was designed for the modern internet era, when most softwares run on private servers and are never distributed physically to a user, all the while powering websites used by billions of people.

I have witnessed sooo many people on the web getting afraid of barely using an AGPL software on their computer. Such a paradox!!! There is a huge misunderstanding on AGPL. The point is: if you don't modify the code, you only have to make the software that's AGPL available to your users.
But if you are the user of an AGPL software (like Signal or Bitwarden), then you have nothing to do. In fact, that's a plus for you as a user. You're using a software over which you have total control. You can see the code and you'll be sure that the source code of any improvement of this software will always be available to you.

It may be the consequence of several articles bashing AGPL, after Google decided to ban AGPL.

Weak copyleft licenses

The two most popular "weak" copyleft licenses are the Lesser General Public License (LGPL) and the Mozilla Public License (MPL).
Their underlying philosophy are totally opposed:

the LGPL was made by the FSF as a compromise to make more people use GPL-licensed softwares (i.e make proprietary vendors use Free Softwares). The copyleft effect is not systematically triggered to all kind of modified work.
the MPL is a file-level copyleft license created by Mozilla to enable proprietary add-on (in particularly made by Mozilla) while encouraging some kind of improvements. It's more of an Open-Source philosophy type of license.

The LGPL is more complicated than the MPL and tends to be less used these days (the trend is simpler, more permissive licenses that are easy to embed into proprietary softwares - the open-source philosophy).
However, the LGPL is more suitable for people who embrace the spirit of the Free Software movement.

More on the difference on my answer to this StackExchange question.

Permissive licenses

Among them, the most famous ones are:

They all share one thing in common. They all say: "use me however you want, even by combining me with other softwares, no matter Free Software or Proprietary Software, as long as you give attribution".

The most praised one these days is the Apache License for its protection against software patents.
There is a huge debate over the very existence of software patents. In fact, many organizations have fought against it, like the FSF, the GNU Project and the Foundation for a Free Information Infrastructure (FFII) which Pieter Hintjens led for two years.
Read Pieter Hintjens for an opinionated essay against software patents. Listen to Richard Stallman 2 hours long talk about the dangers of software patents.

So if you have to choose one permissive license, go for Apache v2 ;).

License compatibility

All Free licenses are completely or partially incompatible with proprietary softwares (because of the copyleft effect).
But not all OSS licenses are compatible with each other.

For example Apache v2 License is infamously incompatible with GNU GPLv2.
This can be very annoying when you are working with some Kernel libraries released under (L)GPLv2. If you're anti-copyleft, then you have no choice but to use MIT or BSD (or equivalent) and you cannot have the patent protection offered by Apache v2.

The following image summarizes the license compatibilities. It was extracted from Wikipedia, credits to David A. Wheeler.

That being said, all the latest versions of the modern most-used OSS licenses are compatible with each other. The real (voluntary) "issue" is the incompatibility of copyleft licenses with proprietary softwares.

Conclusion

Hopefully this article taught you a few things. Remember that choosing a license for a non-trivial software shall never be done lightly. It affects all aspects of your project: its philosophical/ethical stance, its attractivity to contributors/users and its business viability.