[Disclaimer: not a lawyer, this is not legal advice]
I understand they're trying to get rid of cascading attribution requirements for enterprise (which is part of the reason NIH syndrome became so prevalent!), BUT I really dislike this incarnation for two primary reasons:
1. It has conflicting branding. How is a newbie gonna know the difference? How many will misclick it because it sorts first, or because they don't know it's different than true MIT?
2. Some users may read 0 suffix as "original", "optimal", or "first" (0-index), further polluting the branding.
If they rereleased under another name these problems would be alleviated, but I don't think their intentions are entirely altruistic here.
Releasing under a different name would have the opposite problem. No one would use it because they wouldn't know that it was the MIT license with the attribution condition removed, even if that was exactly what they were looking for.
The 0- convention is also somewhat standard for this category of licenses (like 0BSD).
Not a problem in actual practice, derivative licenses are commonly identified as such and their slight differences are much easier to explain than the whole of any one original license.
For what it's worth, my understanding is that the 0BSD license was similarly not authored at Berkeley [1] and, more confusingly, was derived from the ISC license (not the BSD licenses) [2].
That’s correct. Rob Landley and I came up with the same license text but called it different things.
I called it the Free Public License and had it approved by the OSI under that name.
He called it the 0BSD, and argued that since he came up with it first then it should be given his preferred name. I think “0BSD” is an awful name for an ISC-derived license, but eventually I gave up trying to argue with him because I felt that having two names was worse than one bad name.
Yes, but the spirit of "like the BSD license, but go even further and put no conditions on it" is pretty self-evident from the name, as long as you've heard of 4/3/2/1 clause BSD licenses.
I think that is worth something. Something quite marginal, to be sure. But still a nicety.
I have similar concerns about the name being confusing, in the sense that the original MIT license originated from MIT, whereas this license was written by someone else. It might appear as if MIT is endorsing this license despite only serving as an inspiration, and we don't know if MIT actually wants to endorse it.
---
1. It has conflicting branding. How is a newbie gonna know the difference? How many will misclick it because it sorts first, or because they don't know it's different than true MIT?
---
Not a lawyer either and I deeply hate MIT(s) but my favourite concept in law is that ignorance is never a justification and is never used so, if you are using a license, don't be ignorant on that license? Or should we avoid doing things because there is people who wouldn't bother getting documented that could get upset? :o
If you're a big enterprise with lots of open source dependencies, you have to audit your ENTIRE dependency chain to see which of them are MIT licensed, and include the license text for each one.
It's a pain, so many big shops avoid MIT, and more restrictive licenses like GPL
So you're happy to build your commercial product upon the free works of hundreds of FOSS developers, but only if you can't be arsed to give them attribution? Cry me a river! An enterprise has a responsibility to manage its licenses. Complaining that it's not fun is just childish.
I think also it becomes an issue where you have to preclude a source file with ~10+ copies of the same license for each dependencies. It's the way you have to satisfy the attribution requirement (eg "must include this text," rather than "must include a link to this text somewhere in the source tree") that presents the issue.
It's part of why, as I said in the original comment, lots of big shops do _everything_ in house, even when there's a higher quality open source alternative. They simply don't use projects with attribution requirements that are too strict.
Exactly, but I couldn’t figure of how it actually differs from CC0, despite apparent protestation to that effect:
> The CC0 and various "do what you want" licenses and various public domain dedications may be less attractive to the initial developer for various reasons (i.e., a license is preferable to a public domain dedication).
Maybe the difference is more a branding/tribal association than a legal one?
CC0 is public domain. You waive all copyright (and even moral rights!) in a CC0-licensed work. MIT0 only waives the requirement to retain your copyright notice. It doesn't waive your copyright itself.
CC0 contains language that is more suitable for works of art than computer programs, and is intentionally vague regarding which rights are being waived because the use cases are so diverse. (What does it mean to "perform" or "communicate" a program?) Developers might prefer something a bit more precise.
CC0 is long and complicated [1]. You probably don't want to put that in the header of a gist.
> MIT0 only waives the requirement to retain your copyright notice. It doesn't waive your copyright itself.
Except that it does. MIT-0 "only" waives the requirement to retain your copyright notice in that that is the change it makes compared to the MIT license. But the MIT license has already waived every other aspect of your copyright. You're not retaining any rights in an MIT-0 licensed work.
The MIT license doesn't waive any aspect of your copyright and/or moral rights. You still claim to have those rights, however diluted they might eventually become. Otherwise it wouldn't make sense to grant some or all of those rights to other people.
CC0 on the other hand goes like "I don't have those rights. Nobody does." It falls back to the rights-granting model only in jurisdictions that don't allow unilateral copyright-waiving.
In practice, of course, the difference usually doesn't matter. In theory, though, there might be an edge case where it matters, perhaps in situations where the original author's moral rights are involved, or if someone discovers a novel use case for computer code that nobody had imagined before. (Am I allowed to bunch up your code into a virtual snowball and throw it at a virtual alien kid while cursing your name in Klingon?)
Like the MIT license but you must put 2 instances somewhere in your attribution. It is per-license, so the whoever includes your work in their own project must put 4 instances, and so on.
That's only for the case where project A depends on project B which depends on project C which depends on project D...240 projects deep. But what if each of them has more than 1 dependency?
Also you'd have to sum up the entire series, since the licenses at every level are included in the folder, not just the top-level one.
~~Yes, in my head I was considering all of those to be, in their total, the "sublicenses." Should have been better worded; but a total of 240 distinct MIT licensed dependencies in a project is the number at which MIT-2 would run out of atoms to print on.~~
Nevermind, I see your point! In reality it would be something relating to N sums of 2^k, k being highest depth dependency for each of the N projects. Obviously we're talking bounds and not exact calculations here.
This does kind of get me thinking: why do we need to include full copies of licenses in distributions anyway? Couldn’t we just include the name of the license and relevant details?
Oh for the love of God, don’t introduce confusingly named variants of widely known licenses. If a new license is really needed — fine, if lawyers think that’s necessary, it probably is — but don’t name it that. No one will bother to check whether MIT and MIT-7BZ3/2 are slightly incompatible.
I’m just spit balling here, but it sounds like problematic brand association to me. I could see how someone might think that would mean any code written by or for AWS uses this license. Maybe in MIT’s case that’s not as problematic? But tbh, I’m on my phone atm and I don’t know much about the MIT license’s history or origin. Are most developers bad at licenses? I’m feeling pretty naïve right now.
Anecdotal but the majority of developers I run into have no true understanding of a majority of software licenses, if any at all. In my younger years I sat down with a developer who did know and got a summary about literally every license I could think of, and its still stuck in my head since. I feel like colleges should really cover them in some capacity with examples of lawsuits related to licensing disputes.
The MIT license has a bit of a weird history between Expat and X. I spent way too much time looking into this a few years ago. The bottom line is that people were more casual about things in those days.
I prefer zlib over this. If you're shipping my code in a binary I don't care about attribution, but if you're reproducing my entire repo in source code form don't claim that you wrote it
Just curious: why can't they use zlib for teaching? zlib doesn't require attribution and it is a very simple license, so I wonder why the need for MIT-0?
I mean... yes it would, but isn't teaching proper care and attention of licensing important?
If it's not the developer's original code then they ought to think at least a few minutes about where it came from and whether they have consent to use it and all that nuisance.
This license has proven useful for code that is intended for developers to use as reference, teaching samples, examples, or templates that other developers may modify for their own purposes.
I don't think teaching proper care and attention of licensing is incompatible with wanting to release teaching materials under a licence that does not require attribution.
The longer answer is that CC0 is not a licence – it's a public domain declaration. With CC0, you relinquish all rights to the work, including copyright. With MIT-0, you still hold the copyright; you're just allowing people to do whatever they want with the work.
Another difference is the one you alluded to. CC0 is designed for written content, images, videos, etc. It's not really designed for code/software. That's not to say you can't use it for software, however. MIT-0, on the other hand, is designed for code.
So, in essence, these two legal instruments are in a sense equivalent, but there is enough nuance there that it could make a difference in a court of law, depending on how they are interpreted.
There's also the fact that CC0 is quite a mature piece of text at this point, whereas MIT-0 is the new kid on the block. That might be something to consider, too.
Also, obligatory disclaimer that I'm not a lawyer and that this does not constitute legal advice.
> With CC0, you relinquish all rights to the work, including copyright.
With CC0, you relinquish your copyright only, but explicitly retain trademark and patent rights (!).
> 4. Limitations and Disclaimers.
> a. No trademark or patent rights held by Affirmer are waived, abandoned, surrendered, licensed or otherwise affected by this document.
This makes it sort of iffy for software when software patents exist unless coupled with some kind of patent waiver or assertion of not holding and not being aware of patents. While you can argue with MIT-0/0BSD/whatever else there might be that there is an implied patent license, there is an explicit non-license for patents with CC0.
Yes. Patents were why CC0 was withdrawn from OSI consideration. Now lawyers have argued that the patent grant is implied as part being given rights to use the software. But it was sufficiently controversial that many were uncomfortable with it.
I see. I didn't mean to imply CC0 to be used for code in general. But for documentation that has code (reference documentation & code, getting started tutorials, etc).
Which was the usecase for MIT-0 that the previous commenter seemed to hint at.
One issue with this is that there's no obligation to keep the license on it-- which means that the entire disclaimer and lack of warranty can get sheared off the software.
Then in turn downstream users can legitimately obtain it without knowing that you're attempting to disclaim liability.
I think that's not really an issue for the intended use case: "code that is intended for developers to use as reference, teaching samples, examples, or templates that other developers may modify for their own purposes."
Consider some libfoo, and in libfoo you include example.mylang as a "quickstart" for the common use case, and many people use this as a starting point, modifying it a bit to suit their needs but not much. Technically you should attribute this file/function/class too, but in practice no one does, and no one really cares either.
I've had people ask me to put a license on 2-line Stack Overflow licenses (it's already CC-BY, but people still ask for some reason, I guess to comply with company legal dept.)
It's useful for stuff like that: you typically don't really care what someone does with your fairly simple examples/demos/etc. (which may not even meet the threshold of originality to apply for copyright protection) and whether they attribute it.
Think even more trivial than a demo or an example -- think more along the lines of a "skeleton" project. Something that isn't even a runnable example yet, just the outline of where someone else's code will go.
Whether something like this is even copyrightable is questionable. But releasing it under a license like MIT-0 removes any ambiguity -- whether it's copyrightable or not, the original author is allowing you to use it for anything at all, no strings attached.
Consider that code without any license at all, or even explicitly public domain. There is no implied liability at all. It carries as much legal risk as leaving a photocopied Anarchist Cookbook in your sock drawer.
If that's really the case, why does pretty much every software license feel the need to include a liability disclaimer? Genuine question, I've often wondered about this.
I suspect it has more power to prevent litigation (and disputes more generally) than to win anything? It’s also entirely possible it’s just [a much less derisive term than “cargo cult” which eludes me at the moment].
I predict a wave of articles and blog posts about how open source doesn't work in about 5 years, after the new generation of developers that Amazon convinces to use this licence complete the process of being out competed by Amazon using their own IP.
I mean, I’m already pretty cynical about open source. I wouldn’t go so far as to say it doesn’t work, but I think that many open source projects, or at least many of the ones in the fairly juvenile domain in which I operate, have intentions thinly veiled by the guise of corporate altruism.
Is... Is Amazon allowed to just use the name "MIT" for its own unrelated license? Doesn't MIT have some say over the use of the name "MIT"?
Can I make my own license and call it MIT-2, to make my license sound much more established and prestigious than it really is? Or is this a case of "it's OK because Amazon does it and nobody will sue them for it"?
I would think the only way they could be stopped is if "MIT" is a registered trademark. No idea if it is. But actually, would trademarks even apply in legal documents? A software license is not a product nor a business and there's no chance of confusing it with one, so could I publish a Disney™ Public License?
What makes you think that software licenses are not products? Licenses like the GPL or to a lesser degree the Apache 2.0 are products that took a lot of effort to produce and are marketed just like other products. And I wouldn't be surprised if lots of EULAs are basically the same license sold by some lawyer or consultant to multiple companies.
Software licenses don't just fall from the heaven and get adopted on their own.
IANAL, but assuming a jurisdiction in which public domain dedications are likely to be legally valid (such as the US), is there any real difference between this license and a public domain dedication? Isn’t this functionally equivalent to the public domain, even if it is technically not?
Public domain code, I can do whatever I want with it, even publish it with all indications of its origins removed. Legally fine. But it seems like the same thing here-for a downstream consumer at least.
From the original developer’s viewpoint - the warranty disclaimer may give some additional legal protection, but you can put a warranty disclaimer in a public domain dedication too. People will debate whether warranty disclaimers in public domain dedications have the same legal force as those in copyright licenses-but maybe that too is a theoretical debate, because how big is the risk of someone being successfully sued by a code reuser who has zero commercial relationship (support or consulting contracts, etc) with the original developer-has that ever actually happened? Or are all these warranty disclaimers just lawyers saying “it’s never happened, but we can’t totally rule out the remote possibility that one day it might, so let’s just include some legalese to improve our position if anyone ever tries to make it happen - a disclaimer doesn’t cost anything significant anyway”?
Prior work in the public domain can have implications for future licensed work derived from the public domain work.
Releasing under an extremely permissive license does not relinquish copyright. Any future derived work by the author would still be unambiguously licensable under any conditions they see fit.
> Prior work in the public domain can have implications for future licensed work derived from the public domain work.
Isn't that just the normal case of a derived work? If you modify a public domain work, you don't have copyright on the parts which are identical to the public domain original, but you do have it on any non-trivial modifications/additions, and you can license the later (whether under an open source or proprietary license).
A real world example is the Ghidra reverse engineering suite - it was originally written by US government employees in the course of their employment, and as such that original code is ineligible for copyright in the United States and automatically in the public domain in the US (although its legal status in non-US jurisdictions is rather more complex, variable, and at times even unsettled), and hence cannot be released under an open source license (or any other copyright license for that matter). But, they also accept outside contributions, which generally would be under copyright (whether by the contributor or their employer), and it is a condition of accepting those contributions that the contributor agrees to license them under the Apache 2 license. Hence, Ghidra is an amalgam of public domain code (or at least, public domain in the US) and copyrighted code licensed under Apache 2. I'm not aware this causes any problems in practice.
That seems pretty strange to me. Even if the contributions by themselves are not PD, they're inextricably linked to a PD work in a way that makes them lose meaning when removed from it. And they're still being incorporated into the work by a government employee, which means that the work as a whole must still be PD, even though small parts of it might not be.
That's a great question. The short answer – no one knows. Technically someone does own the copyright to these words, and if they are able to prove it they could sue every project that uses or refers to it, which is a troubling thought. That is very unlikely to happen, however, since the license originated at MIT (as the name suggests) and the current version is the result of a lot of iteration and modification done by many people over many decades.
And trademark probably plays in too IANAL. I assume coming up with your own Creative Commons license unrelated to that organization would be problematic.
> Technically someone does own the copyright to these words, and if they are able to prove it they could sue every project that uses or refers to it, which is a troubling thought.
That seems unlikely to be true. Recipes can't be copyrighted because they're functional. Legal licenses are also functional.
(What's that? You think source code is functional? Shows what you know.)
No, the licenses themselves are protected by copyright as well. But the only license I can think of that explicitly states this fact is GPLv3, which begins:
So MIT is "do attribute me" and I guess MIT-0 is "you don't need to attribute me (but it's fine if you do)", I wonder if there are any "explicitly do NOT attribute me" licenses?
I use this license for a few of my NES libraries, since they're meant to be used in homebrew games where memory is a real problem. 32kB of PRG ROM (and the associated 8k of CHR ROM) doesn't leave much extra room for text rendering systems or fonts, and every byte counts. Sure if there's room, attribute away, but for simpler games there's really no need. It's fine :)
Didn't realize Mark decided to do this after leaving HP Enterprise, but crazy that it took 4 years to get published by Amazon.
For an open source evangelist, this is an odd way to engage the community. Really makes the FSF's principles look good, with the four freedoms and such.
> i.e., a license is preferable to a public domain dedication
Are we that easy to fool by lawyers? The public domain is the public domain. Before someone comes out and says that the concept of dedicating IP to the public domain doesn't exist in some countries or is full of hidden perils, I'll ask you to provide a single case where code was attributed to the public domain and that became a problem for the author or for the users.
It relates to availability, not copyright. “Not secret” instead of “not copyrighted”.
So if you say “the facts are in the public domain” in BrE, that might mean that they’ve been published in a newspaper. It doesn’t mean anything about their copyright status.
Some people do want to simply put their work into the public domain to the degree they can.
Personally a permissive license probably makes more sense but I appreciate someone saying they don’t want to have anything to do with this copyright stuff.
I have used CC0 for code that I would like people to be able to use in code licensed however they choose, without having to bother with attribution. Sample code, for instance.
If they're being responsible, they only have to note the provenance of the code in version control. After that, they can basically treat it like their own.
The MIT license isn't owned or managed by MIT in any capacity. The name was given to it many decades ago (since it is derived from a bunch of licenses used by MIT at the time) and it just stuck.
> This is more like work for for free for Amazon license
Pretty much any open source license would be that, the main difference this one provides is that Amazon wont even have to acknowledge you worked for them for free :-P
I believe it is. Note that 0BSD is not related to the other BSD licenses, but is rather textually derived from the ISC license. The ISC license is pretty much just "what if we took MIT, and removed all the text that is already implied in most jurisdictions?"
So 0BSD is to MIT-0 as ISC is to MIT. Just a less verbose alternative.
There’s also been the unlicense fairly recently. They’re all intended as lightweight ways to effectively put something into the public domain—to the degree you can in a given jurisdiction.
And there’s also the anti-copyright ideological baggage associated with unlicense. To round it out there’s also CC0 but OSI objected to the explicit lack of patent grant language even though patent grants are often assumed.
Isn’t this effectively the same as WTFPL? Of course I’m not a lawyer but they both seem to have total overlap in purpose and effect. The disclaiming aspects of MIT are basically restating defaults anywhere they’re legally enforceable.
In practice, the wording is much more legally problematic in the WTFPL. And licenses such as MIT-0 or 0BSD contain liability waivers that are not present in the WTFPL
There are tangible costs to crayon licenses, long before cases reach court. The more convoluted and amateurish the drafting, the more unclear the meaning and the more that lawyers make advising clients what to do about them.
Sticking it to users who care about compliance is part of the appeal of licenses like the WTFPL, but it's ironic that the money flows to lawyers.
When the ASF finally banned the JSON license for use in dependencies, the legal cost of figuring out compliance was one of the motivations cited by the people who proposed the ban.
For the GPL, this makes sense. In rental contracts this can be incredibly annoying. I’m one of those people who reads them from start to finish, and very frequently I have to have them amended because they’re obviously either:
- reused from previous rental terms, incorrectly restating basic local laws which have since been revised
- downloaded from some website, with terms which are obviously illegal locally
These contracts just introduce more ambiguity. They really should just reference the relevant legal codes. And at least in places with good protections for renters, that should maybe be mandated so the actual lease document is reduced to meaningful terms the renter needs to know.
Now I want a MIT--1 license that prohibits attributing the code to the author in any way, including in legal filings, and then you don't need any other part of the license.
At that point just go fully PD or CC0. Why have some random license nobody has heard of or wants to read about when you can just slap a CC0 sticker on it.
> a license is preferable to a public domain dedication
Who says that and why? Why should I care about the complicated terms of a license - especially when such a license expresses essentially the same as "public domain"?
Or in other words: is there any real rationale behind the above statement?
What happens once there are layers of execution of the lack of attribution, but then someone downstream comes into a battle over provenance? They have to backtrack now even more poorly documented provenance in discovery? Sounds awful.
Is this sarcasm? This license is horrible for developers.
Attribution is literally the only thing they get out of open source. Any open source developer who uses this license could not possibly be using it out of their own free will.
I understand they're trying to get rid of cascading attribution requirements for enterprise (which is part of the reason NIH syndrome became so prevalent!), BUT I really dislike this incarnation for two primary reasons:
1. It has conflicting branding. How is a newbie gonna know the difference? How many will misclick it because it sorts first, or because they don't know it's different than true MIT?
2. Some users may read 0 suffix as "original", "optimal", or "first" (0-index), further polluting the branding.
If they rereleased under another name these problems would be alleviated, but I don't think their intentions are entirely altruistic here.