Comment by tannhaeuser | Hacker News Reader

Comment by tannhaeuser | Hacker News Reader

Good article, but I thought webpack is founded by The OpenJS Foundation these days, who are making an astonishing job of identifying and sustaining core web packages. Unfortunately, they also picked controversial software like AMP (in exchange for contributions by Google I guess) to make it appear a grassroots or "legitimate" effort. Also, I think the ongoing trend to slap a thin wrapper on top of F/OSS and sell it as a service (aka "the Cloud") points to a weakness in F/OSS licensing schemes that needs to be addressed as I'm sure that almost all devs working on F/OSS don't intend their work to become part of a lock-in scheme, mass surveillance, or both, all the while not even getting contributions back. Apart from the licensing situation, F/OSS developers serious about generating an income might also need to work on integration and polish; at least it's what the success of stuff like Docker suggests, which, as super-practical as it is, in a way is benefitting from "too much" F/OSS, too many Linux distros, either too much or not enough NIH in F/OSS, F/OSS failing to directly engage with their user base, and/or developers not personally promoting their software such that the whole thing is falling prey to large corporations giving the illusion of order and maintenance, when the reality is that the initial development of any package of significant depth and its maintenance can only be provided by individual developers or small teams in the end.

> Also, I think the ongoing trend to slap a thin wrapper on top of F/OSS and sell it as a service (aka "the Cloud") points to a weakness in F/OSS licensing schemes that needs to be addressed as I'm sure that almost all devs working on F/OSS don't intend their work to become part of a lock-in scheme, mass surveillance, or both, all the while not even getting contributions back.

the solution to "not even getting contributions back." exists since 2007 but apparently just writing the letters "A" "G" "P" "L" is enough to make a lot of people screech.

The AGPL doesn't guarantee that you'll get contributions back.

For the very few companies who are willing to touch AGPL code, they only need to make the changes available (e.g. a CD in the post is sufficient).

Very few companies are willing to use it, because it creates an almost unassailable hurdle in lots of industries.

It's also highly at risk at making a bunch of other projects AGPL as well.

<edit>

Using "you" in the general "one" sense, not to attack you personally of course

</edit>

> it creates an almost unassailable hurdle in lots of industries.

It's intended to be a hurdle, if you want to build a business on other people's work and not contribute back. Be a player that benefits the ecosystem or go away and build the product from scratch if you think you can.

> a CD in the post is sufficient

Yeah, it would be. If they make valuable contributions to the source code, they'll soon realize that just dumping those on github will have the same effect and save them a full time employee for burning CDs. If the contributions are not valuable, the problem will likely solve itself when the company goes bust.

It does nothing of the sort. This is propeganda that mega-corps want you to believe because they are unable to rip off AGPL software effectively. Organizations like Google can afford to not use AGPL software (they can just write their own software), but it costs them money, so they refuse to use AGPL software and spread propeganda that it's impossible to do so, to weaken its market penetration and ensure that they have greater access to software without the responsibility to contribute back.

In actual fact, being compliant with AGPL software is extremely easy. Nothing at all is expected of you if you just ship it into production as-is. And if you modify it, the only obligation is to send the modifications upstream as patches. That's it. Surely any company is capable of completing such a trivial feat.

You're glossing over the fact that being compliant with AGPL requires more competence than the vast majority of software jobs. I know you could competently integrate or use AGPL software, and I might be able to do it (while repeatedly referencing armchair lawyers on stackoverflow), but there is still a risk I fuck it up and my (or my employer's) competitive advantage is in jeopardy.

You are right that AGPL is easy if you can just run it as-is. The real problems start with libraries. If I link an AGPL PDF-generating library with my monolith service which runs financial reports over my business, what happens? Do I need to now publish my SQL queries that access sensitively named tables? Maybe that reveals new secret projects we're working on, or perhaps security incident mitigations underway.

Sure, there are countless ways to architect my way out of those problems, but what if I don't have the resources to do that right now?

The small obligation becomes a large burden pretty quickly.

(The real solution to the above problems is to pay for the corporate license that removes the AGPL obligation, but that assumes you can afford to do so)

AGPL is fundamentally scary to lawyers and management because they're afraid someone will sink the ship. Sure it's mostly FUD, but there are some serious valid concerns.

>If I link an AGPL PDF-generating library with my monolith service which runs financial reports over my business, what happens?

No one uses AGPL for libraries, except maybe by mistake. LGPL is used for this purpose.

>Do I need to now publish my SQL queries that access sensitively named tables?

No. The FAQ from GNU makes how this works pretty clear:

https://www.gnu.org/licenses/gpl-faq.html#AGPLv3InteractingR...

The AGPL is not viral in the sense that all client software accessing an AGPL-licensed service are required to use the AGPL, but rather, such clients are entitled to the source of that service. It puts no obligations whatsoever on client software.

https://www.gnu.org/licenses/gpl-faq.html#AGPLv3InteractingR...

Maybe you should actually read up on the AGPL license before you make assumptions about it? It seems like you don't really understand it.

> No one uses AGPL for libraries, except maybe by mistake.

This isn't my experience at all. Here's an example:

https://github.com/unidoc/unipdf/blob/master/LICENSE.md

> The AGPL is not viral in the sense that all client software accessing an AGPL-licensed service are required to use the AGPL, but rather, such clients are entitled to the source of that service. It puts no obligations whatsoever on client software.

This depends on what your "client" is. If you're making library calls, lawyers get uncomfortable.

> Maybe you should actually read up on the AGPL license before you make assumptions about it? It seems like you don't really understand it.

I understand it quite well, and I even use it myself! There are just problems associated with it that most proponents gloss over.

>This depends on what your "client" is. If you're making library calls, lawyers get uncomfortable.

Ah, in this case, this software is designed to maximize conversions to the paid commercial license, so they deliberately use the AGPL in a way which makes it inconvenient for your internal use. This is not common in software which does not have an alternative commercial license. I don't really appreciate this model because it is disrespectful of the copyright of third-party contributions.

I agree with you, and this is where my concern lies: it's Open Source software, but I think that terminology is misleading (in the case of this software), where as, I find some of the close-but-no-cigar licenses much more "Open". So, while I see tremendous value in OSI's definition, I find it is somewhat of an impedance mismatch with what the average fool expects.

> No one uses AGPL for libraries,

I actually do. If I write a library, I definitely don't want people to put a web UI on top of it and get away without any obligations, like so many people do for e.g. website that are just ffmpeg frontends.

> the only obligation is to send the modifications upstream as patches

That is incorrect, the modifications have to go to users not upstream. All the GNU copyleft licenses are like this.

Sorry, you're right, thanks for the correction. In practice, sending the patches upstream is the easiest way to be compliant, though.

You're probably right (and also, I find your comment further below regarding interpretation of AGPL insightful, insofar as it touches on a point in AGPL that would impose limits onto unrelated code, and that some people consider unenforceable).

Just wanted to mention that I believe AGPL is sometimes considered with a dual-license scheme in the hope that enterprises buy the commercial license, or to avoid a situation where a commercial third party benefits from a piece of software disproportionally, or to the detriment of the project's funding (such as providing pure hosting), whether contributions are upstreamed or not. Not being able to express clearly what you're after - cash or code - makes AGPL suboptimal. It's a reality that software without funding and maintenance only goes into bitrot mode, so why not come out straight and sell potential customers maintenance and support? Clearly, the software licensing universe needs additional considerations today compared to the time when Affero/AGPL was conceived.

> almost all devs working on F/OSS

... a large fraction of OSS work these days is sponsored by corporations.

To clarify/correct this a bit, Webpack is not funded by the OpenJS Foundation. Yes, it is an OpenJS project, but the foundation does not provide financial support for open-source development. It does provide legal, marketing, and other support, as well as being able to assist with infrastructure costs, but it's still up to each project to separately run (or not) their own fundraisers and other funding activities.

Also, it would be incorrect to say that the foundations "picks software" such as AMP; instead, projects apply to join. The decision to accept AMP was made entirely without consideration for Google's status as a sponsor of the foundation. And finally, while it may count as splitting hairs, Google retains ownership and control of the AMP cache; it's the core technologies that have been transferred to the foundation, not the instance run by Google that relies on those technologies.

Reply on news.ycombinator.com