Hacker News Reader

> Why is it that there is no standard applied to those who publish code for distribution purposes?

Because it's rude to make demands of someone who is doing you a favor.

Because a system that adds costs to profit-free work will collapse.

Because your "distribution" line-in-the-sand doesn't exist. I assume you're thinking of NPM or pypi, but ex. Debian doesn't ask people before including their packages, and ex. nixos pulls directly from those "non-distribution" channels.

> My point is that they don't legally owe anyone anything but we should impose a moral standard in lieu of a legal one. If you are saying "here's this code, I've packaged it up and sent it out for distribution" I think it should be perfectly fine for us to say "did you do the bare minimum to make this code acceptable for others to use?".

Okay; let's also make a moral standard of paying people when we derive value from their work. I think it should be perfectly fine for us to say "did you do the bare minimum to repay the person who gave you this code to use?".

> I don't get why we say "you have no ethical obligations in open source", why do we do that? Who benefits? I get not having legal obligations, but once you're distributing code for use it seems absurd to say that you have no ethical obligations.

We do that because we benefit. Making things easy for the people who are giving their work away for free helps foster an ecosystem where people keep giving stuff away for free.

> You chose to do that, you chose to distribute it, you didn't have to do that.

Yes, that's the point. We'd like people to keep giving things away even though they don't have to. If you try to impose costs on them for doing that, you'll alter the incentives so that they do the rational thing and stop giving stuff away, and/or start charging for it.

> Debian doesn't ask people before including their packages, and ex. nixos pulls directly from those "non-distribution" channels.

Then Debian (or NixOS) are publishing the code for distribution, and Debian (or NixOS) should be morally obligated to do the bare minimum to make the code acceptable for others to use.

no

> Because your "distribution" line-in-the-sand doesn't exist. I assume you're thinking of NPM or pypi, but ex. Debian doesn't ask people before including their packages, and ex. nixos pulls directly from those "non-distribution" channels.

Correct, I'm talking about publishing to a package index. The fact that the line is murky right now isn't important, that's exactly the sort of thing we should be clarifying and changing.

> Okay; let's also make a moral standard of paying people when we derive value from their work. I think it should be perfectly fine for us to say "did you do the bare minimum to repay the person who gave you this code to use?".

I don't think it's an "either/or" situation, the two are not exclusive. I would like to see more funding for open source software, but that doesn't solve the fundamental issue. It's also extremely hand-wavy. Do I pay you a monthly stipend to implement 2FA?

It also assumes that open source developers aren't getting "paid". They are. Github is free, PyPI is free. You are using services for free, that is a form of payment, or at least a social contract between various developers. So it should be reasonable then to negotiate that contract, eg: PyPI saying "if you use us as a distributor of your code, we need you to enable 2FA", which they now do.

> We do that because we benefit.

Who's we? Because a lot of people don't benefit. When there's a supply chain compromise because a developer used a weak password and no 2FA people are harmed.

> We'd like people to keep giving things away even though they don't have to. If you try to impose costs on them for doing that, you'll alter the incentives so that they do the rational thing and stop giving stuff away, and/or start charging for it.

Of course, and I'd like to keep those incentives as much as possible. I think there is a very happy medium here that isn't strictly "if I publish open source code for distribution I can do no wrong whatsoever, I have no obligations whatsoever". PyPI and other distributors now enforce that if your code gets to be very popular you must enable 2FA. That is a cost, but it's a very minimal cost targeted to a small group.

What I'm advocating for is that we standardize a set of responsibilities as an industry for those who distribute code. They can be very minimal and still have massive impact.

The alternative, in my opinion, is that we'll see increased regulation, because these problems aren't going away and they're going to get worse if we don't collectively try to improve.

Do you demand that every screwmaker make aircraft-grade screws? Aircraft makers need screws and it would be very convenient to them to be able to go down to any hardware store and just buy whatever screw they want since they are all up to spec. No need to evaluate their suppliers since everybody is required to make things up to their demanding standards.

The problem with this is that not everybody needs expensive aircraft-grade screws. Most people only need hobbyist-grade screws, or construction-grade screws. The requirements depend on their usage and it is up to the consumer to correctly identify their requirements and use the appropriate product that is fit for purpose.

The problem with software today is the rampant, careless usage of hobbyist-grade dependencies in critical software. It is the fault of the entities including dependencies that are explicitly hobbyist-grade or inadequate for purpose that poses a problem. It should not be the responsibility of makers of hobbyist-grade screws to produce aircraft-grade screws because the aircraft makers want to go to Home Depot and pick out whatever screw is the cheapest.

The solution that matters today is holding the consumers of these endless software dependencies to task for the usage of substandard or even defective software components, whether open source or proprietary, like every other industry where you must use suppliers that are fit for purpose. To demand a change to the software that is explicitly marked as unfit for purpose is to solve the problem of a aircraft manufacturer using screws from Home Depot by making Home Depot required to only stock aircraft-grade screws while demanding they keep the prices the same.

I'm quite sure that screws are indeed something that, if made in a shoddy way, would have legal repercussions for those producing them. That's the norm - it's software that's weird for not having that. I would actually expect screws are even rated for specific work. Also, people sell screws, so the analogy really makes no sense.

I'm not suggesting that software developers be required to do anything if they're just writing code, and I'm not suggesting that they do things to some sort of extreme, as you seem to be implying, if they do distribute their code for use.

> explicitly hobbyist-grade or inadequate for purpose that poses a problem

Well, no, they're not explicitly hobbyist-grade. That would be fine if someone were just publishing code and saying "don't use this", but they are publishing code for distribution to package repositories. They don't have to do that, they could just leave it as open source code that isn't distributed, and note that it's not production quality explicitly as you suggest.

Perhaps a more appropriate analogy would be if you were making dinner. You go to the farmer's market and someone with a booth their says "I'm giving away some free fruit, here you go". You would hope that someone who set up a stand at the market would be giving you fruit that's edible. If you went home and ate it, and then you got sick because it didn't meet food quality standards, you would not be the one liable, the vendor would be. "But the vendor gave it away for free!" Yes, but other than software the person giving you something is in fact liable for its quality.

Anyway, analogies suck, I'm sure this misses plenty of important bits. Rather than argue about analogies, let's clarify the actual argument.

1. No one forces anyone to publish their code for distribution purposes

2. When you publish code with the intent for others to use it there should be an obligation to provide basic quality standards to avoid that code doing others harm

Package repositories were never intended nor designed for anything other than hobbyist grade software. The have never offered any guarantees on the quality of their contents and in fact the vast majority of their contents are freely accepted without review and explicitly disclaim any fitness such as via Clause 15 of the GPLv3[1]:

"THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM “AS IS” WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION."

If the quality of a package is not positively asserted, then it must be assumed to be hobbyist-grade regardless of how others wrongly treat it, how convenient it would be to assume it is not, or even if the quality is high but no explicit positive guarantee has been made to that effect. It is ultimately the responsibility of the users of a package to verify it is fit for purpose which demands either a positive, legally binding guarantee from the package creator or a assessment of the package itself (which would be easy if the quality is high, but no explicit guarantee was provided). To use a package while expecting guarantees on its function when it explicitly disclaims any guarantees is to use the package wrongly no matter how many others may do so and it is a problem with the user of the package, not the creator of the package who accurately advertised the quality of their package. It is the recognition of this fact and the usage of components that have actual guarantees about fitness for purpose that distinguish civil engineering from software development, not the quality of the underlying components; the quality of the underlying components follows naturally once liability is accepted and guarantees are required.

[1]: https://www.gnu.org/licenses/gpl-3.0.en.html

no.

> Why is it that there is no standard applied to those who publish code for distribution purposes?

What does "publish code for distribution purposes" mean? That sounds like all published code.

Does that mean that I can't put my own hobby code up in public? That's a sure-fire way to kill the community dead.

> You chose to do that, you chose to distribute it, you didn't have to do that.

I think the burden is more properly on the people who choose to download and use it, knowing what it is.

> but we should impose a moral standard in lieu of a legal one

I agree with you, but these moral obligations tend to get enshrined in law eventually (or quickly! See Covid)

They're gonna get enshrined into law eventually one way or the other. If we do it ourselves and we're effective at limiting the damage we cause we'll be able to maintain control over our own processes. If we don't it will be taken out of our hands.

> If we don't it will be taken out of our hands.

Which will take the code out of theirs. Disincentivize sharing, and sharing goes away.

> Disincentivize sharing, and sharing goes away.

And existing proprietary software companies will rejoice, as the barrier to entry in the software market will again be very high.