Hot take, don't @ me: AGPL should never have been an OSI approved license.
The license is basically a compliance trap, and I would posit that > 99% of modified AGPL software violates the license.
This leads to a world where license enforcement is entirely up to arbitrary decisions by the copyright owner. Any contributor to an AGPL software package (without CLA) can go and mass-send C&Ds and/or request $$$, kind of like the Creative Commons copyright trolls have been doing for CC-BY violations.
That's without even getting into the fact that really, the AGPL is GPL + a "usage restriction" clause camouflaging as not being one (because that would be against the FSF's principles).
I suspect most of the AGPL users would be better served with some of the more permissive source-available licenses. Unfortunately, people seem to have allergic reactions to non-OSI approved licenses, and AGPL being OSI approved even though it sucks means it gets regarded more highly than better alternatives.
@marshray @delroth I don’t mind using an unmodified version for personal use, but I will not, under any circumstances modify it. If my change isn’t accepted upstream then I have a potentially growing liability and that just isn’t worth the effort. I may grumble on issue trackers, but if you release your code under AGPL then you will never see a single one-line bug fix from me.
If it’s under a permissive license, I’ll happily add new features, test them, and upstream them if I think anyone else might care. Doing so comes with no legal liability and gives me almost the same rights over the resulting project as I have over my contributions.
@PJB @gsuberland @delroth This is why i went with 3-clause BSD for most of my projects.
It lowers the chance of some big corp from successfully suing me over bugs and gives me a chance of getting credit for my work, but stays out of the way for most actual users.
@gsuberland @delroth Indeed. Watching said larpers try and "re-license" a project without a CLA or near-unanimous contributor consent generally makes me go "you do know that's not how this works, right?"
Then there's people that try to attach additional clauses to the GPL, which unless they're exceptions to the license (such as "things using this defined boundary should not be considered derivative works for the purposes of this license") are utterly unenforceable given the license contains an implied right to "revert" to the unmodified version when distributing derivatives...
@delroth my general view on the matter is that since the three options are:
1. they will just not use your code
2. they will use your code in violation of the license and you can't do anything about it
3. they will comply with the license by putting your license text in the installer somewhere (where nobody will read it) and then zip up your code and put it on their website for download, which is functionally no different than getting it off git
... it really makes no odds either way.
@delroth @sleevi
What's the usage restriction in the AGPL? Doesn't it basically say that users of the software have the right to modify it whether it's delivered to a computer in their possession or offered over a network connection?
There are criticisms to be made of the AGPL, but I don't see how to characterize it as a "usage restriction".
"I suspect most of the AGPL users would be better served with some of the more permissive source-available licenses."
I'm curious what you mean here. By users presumably you mean of the license rather than users of the software. Is that right? If so, what are you thinking of as the goals of AGPL licensors that would be furthered by source available licenses?
@delroth OSI approved the OpenWatcom license and it does what AGPL is trying to do more explicitly.
Ok, it still has some severe problems, but it's a lot more legally ironclad to say "you must publish private forks" than "you must implement a view source URL that may or may not get stripped out by a proxy".
Bonus points: You get to be Open Source but NOT Free Software! Perfect for running on GNU-less Linux distros like Alpine Linux.
@delroth left without commentary ;)
@delroth
I get that perspective, but respectfully I think it has a place in the OSI. The GPL limits some individual freedoms in order to support grander community freedom; a noble but sometimes questionable goal. I think the AGPL has the same spirit.
It is definitely a compliance trap for the unaware.
@bagel and the previous part of the sentence is even more key imo:
"if you modify the Program, your modified version **must** prominently offer all users interacting with it remotely through a computer network [...]"
(emphasis mine)
(no @ here, just FYI for anyone coming to this thread)
EUPL is a thing and OSI-approved, although it's more comparable to the MPL than AGPL. At least in the EU, the EUPL is definitely enforceable, and all 22 languages are considered official, legally binding versions of the EUPL.
https://commission.europa.eu/content/european-union-public-licence_en
@delroth I'm particularly frustrated at the way the unclear nature of the license is used as a mechanism to filter out users. Don't do that! Just clearly write what you actually care about and let people make an informed decision.
Contracts (and thus licenses) aren't arcane incantations: there should be mutual understanding at their heart!
Pierre Bourdon
Marsh Ray
Nasado
David Chisnall (*Now with 50% more sarcasm!*)
Ellenor Bjornsdottir 
Graham Sutherland / Polynomial
PJB
Andrew Zonenberg
Josh Simmons
Emelia/Emi
Drew 🐘
Ian McKellar
waldi
aurorus
Haelwenn /элвэн/ 
Dark Sage Torunka 
Irenes (many)
Diego Elio Pettenò
luna the doggie 
Mathieu Fenniak
ScreechingBagel
Alfred M. Szmidt
aura-deprecated
Niklas Korz
Alice I. Cecile
HJVT