If you arrange to give people the same money you would have paid for the closed sourced solution maybe they will let you keep your modifications to yourself and sell the resulting product.
For example just running AGPL software anywhere on Google's servers would open them up to having to open source the entire of Google search. That's why Google has banned the AGPL: https://opensource.google.com/docs/using/agpl-policy/
Google developped an excellent library that they open sourced called Google OR Tools [1]. It not only provides access to Google's in house linear programming optimiser, but also acts as a wrapper around other solvers, some free, some commercial, so that you just change a parameter to try another solver.
However because of conflicting licenses or licenses restrictions, google cannot distribute binaries that include certain FOSS solvers, like GLPK. So to be able to use GLPK you need to recompile the whole project yourself. If you go on the Google OR tool forum, you will see that it's not a simple CTR+SHIFT+B, everyone is struggling, and I wasted a lot of time trying to do this.
That's what I mean by strings attached.
This is misinformed fear mongering (or maybe FUD?). The AGPL only requires that you offer the source code to the "Corresponding Source"[1] of the AGPL licensed program, if you modified[2] it, to anyone using the program over a network. Merely using an unmodified AGPL licensed program doesn't require anything. Just like other GPL-family licenses, the AGPL is not an "EULA" and doesn't apply to simply using software.
The only way Google's search software could be affected is if AGPL licensed code was added, which is why they have (with good reason) a policy to not do that. Similar risks would apply to including proprietary code, too.
The entire definition of "corresponding source" in the AGPLv3 is identical to the language from the equivalent section of the GPLv3 (and it also includes the same "based on" language and section on aggregates) which is intended to apply to dynamically linked libraries, and it has been widely interpreted that way. If the software would be a single work as defined by the GPLv3 when distributed on a physical CD then it would be covered by the AGPLv3 when accessed over a network.
> the AGPL is not an "EULA" and doesn't apply to simply using software.
All GPL family licences are effectively EULAs because it's the GPL licence that grants the user permission to run the software, without that permission the user would be violating copyright by copying the program into RAM to execute it
>The only way Google's search software could be affected is if AGPL licensed code was added, which is why they have (with good reason) a policy to not do that.
The Google policy prohibits use or even installation. From the link:
>Do not attempt to check AGPL-licensed code into google3 or use it in a Google product in any way.
>Do not install AGPL-licensed programs on your workstation, Google-issued laptop, or Google-issued phone
As far as I know Google uses a single repository for nearly everything, which is why they are so keen to avoid AGPL code coming anywhere near it. Using AGPL code at Google necessitates "adding" it to the single huge google repo.
If you want to know why they support a license that violates freedom zero you will have to ask them but to me it suggests that the foundation of Free Software is fundamentally flawed in a world of networked software: i.e. the four freedoms are not compatible. My right to exercise freedoms 1, 2, and 3 limits your freedom to exercise freedom zero. Currently software authors have to pick one side of that argument either by using AGPL (restricts freedom 0) or GPL (doesn't protect freedoms 1, 2, or 3).
Please stop spreading FUD about Free Software. The AGPL does not restrict usage. See Section 13 - the requirement to offer to network users only applies if you also make a derivative work by modifying the software, and even then it only requires you to offer the corresponding source to the work that was modified with the new modifications.
This is from the legal analysis by Eben Moglen & Mishi Choudhary for the Software Freedom Law Center[1]
>Whether to extend the copyleft concept to the delivery of services by free software over a network is a complex issue long discussed in the free software community. Freedom zero requires that any user be allowed to run any program for any purpose, which of course includes the provision of computing services to others. Freedom two requires respect for the right of private modification. Their combination requires that anyone be able to run privately-modified copies of GPL’d programs for the purpose of providing computing services to others.
That's very clear that freedom zero and freedom two necessitates allowing people to run modified GPL programs on servers without making the code available to users. The AGPL is explicitly designed to prevent that, so it's inevitably in conflict.
>even then it only requires you to offer the corresponding source to the work that was modified with the new modifications.
As I demonstrated in my other comment the language in the AGPLv3 is identical to the GPLv3 when it comes to the issues of derivative works/combined works/aggregate works. Everybody accepts that the GPLv3 requires software linking to a GPLv3 library to be made available under the terms of the GPLv3. The same applies to the AGPLv3, so in many cases it will mean people will also have to release their own code that links to the AGPLv3 library they are running on their server, if they make modifications to the AGPLv3 code.
[1] https://www.softwarefreedom.org/resources/2014/SFLC-Guide_to...
In the real world people typically want to use code as part of their own service/application which contains domain specific logic they wrote themselves.
>You can perfectly well run an AGPL webserver and not be forced to put your operating system under AGPL.
Because the GPLv3 and AGPLv3 (the licences are identical in this regard) have explicit sections excluding operating systems and system libraries from the terms of the licence.