8

This is a thought that has occurred a number of times for me, especially when concerning questions that asks for help deleting files. Most recently with question 299793 today, which asks with help to identify and remove folders that only contains one folder with a certain name.

In this particular case, the user states that he's well aware of the "risks and implications", but I nonetheless found myself typing "I take no responsibility for the loss of data" at the end of my answer.

Do we have any type of disclaimer to refer to (that may be implicit) when it comes to applying answers that unintentionally (through either a bug in the answer, or by applying it in the wrong way) may result in the loss of data? Or are we personally responsible if our submitted answers are faulty?

Obviously, answers should be properly tested (and ideally, users should use their brains before copying and pasting a command from a web site into their terminal), but there will always be cases where something is overlooked.

I found no such disclaimer in the U&L help pages.

Note: No answer that I have seen or written has caused any damages, as far as I know.

Update: I found this in the "Warranty disclaimer" in the Stack Exchange Network Terms of Service:

Much of the Content of the Network is provided by and is the responsibility of the user or subscriber who posted the Content.

I assume this means I'll be adding more disclaimers to my answers...

Update: Similar concerns have been raised in the past for other (albeit non-technical) forums on Stack Exchange.

And at least one other forum has a disclaimer about not offering "personalized, professional advice". The reasons are obviously entirely different though, but our users may unwittingly also relate to the author of an answer as a source of near absolute authority, based on their high "reputation" and skill at explanation.

Image from the judaism.stackexchange.com forum

5

IMHO questions about deleting files and formatting partitions and creating filesystems and installing a 2nd OS all invite the possibility of accidental destruction of desired data, and so it's caveat emptor. We are not pulling the virtual trigger, and have to rely on the asker's interpretation of reality. It's up to them to gain the knowledge, confidence, and backups necessary.

That being said, I'm a fan of answers / scripts that echo out destructive commands before choosing to execute them.

  • 1
    echo rm -rf "$directory" in non-trivial loops is a good practice I'm already adhering to (even for myself). – Kusalananda Aug 3 '16 at 6:09
  • @Kusalananda This exact code is horrible practice. If you want to display commands that are getting run, use set -x instead of something that will become inaccurate over time as people update the code that gets executed but not the message that gets displayed. If you want to prompt the user, 1. don't do that, 2. don't do that, 3. write English and not code. – Gilles 'SO- stop being evil' Aug 12 '16 at 15:43
  • 1
    @Gilles The idea was to prevent the rm from running at all, and instead use echo to display what should have happened. After running it once or twice with echo (to make sure rm would do the right thing), one would remove it and have rm actually do its thing. Is this bad practice? – Kusalananda Aug 12 '16 at 15:48
  • 1
    The "run-with-echo, confirm results, edit-out-echo, then execute" cycle is what I was referring to. – Jeff Schaller Aug 12 '16 at 15:51
  • 1
    @Kusalananda It's usually bad practice, yes, due to confirmation fatigue. Asking yes/no questions is not good as a safety feature because people soon switch to “yes, sure, of course” mode. The best safety feature is being able to undo the action. When this isn't possible, only confirm things that are both important and unexpected — and work what you're going to confirm into the confirmation. If the purpose of the script is to delete that one directory, don't ask “do you really want to delete [name of the directory that was passed as a parameter to the script]”. – Gilles 'SO- stop being evil' Aug 12 '16 at 15:57
  • 1
    @Gilles I agree with everything you're saying, but this was meant for visual inspection, a sort of dry-run, not for providing interactive confirmation. This comment thread is a bit of a side-track. All I wanted to say was that I also think that the "dry-run-with-echo, confirm result, execute" cycle that Jeff alluded to was a good idea. – Kusalananda Aug 12 '16 at 16:15
4

The disclaimer is present in the terms of service.

The Services, Content, Network and any Software are provided on an "as is" basis, without warranties of any kind, either express or implied,

There's a lot more similar language that only concerns Stack Exchange's liability and not content authors.

There are a couple of Stack Exchange sites that have an additional disclaimer, e.g. Law (text linking to the disclaimer page).

I guess a tweaked version of that could be put in the help section of every site. Out of the way — not on every page. We don't need more useless verbiage on every page.

If someone doesn't use their brain before copying and pasting, they won't pay attention to disclaimers either.

Answers should of course make it clear when a command is dangerous. When the question is about deleting file, it is obvious that answers will show how to delete files, and people copy-pasting blindly risk deleting the wrong files. No amount of additional disclaimers is going to deter the blind copy-pasters.

2

This:

not offering "personalized, professional advice"

is actually relevant for us on U&L.

We can divide today's UNIX systems into two categories:

  1. AIX (IBM), HP-UX (HP), Solaris (Oracle): Which are supported by their vendors. And (as far as I am aware) neither IBM or Oracle people come here to U&L to officially support their systems. It may be the case where an employee of Oracle working on Solaris comes here and answers questions about Solaris, but he is not being paid by Oracle to do so (Disclaimer: at least not as far as I am aware :) ).

  2. Linux, *BSD (including Suse, RedHat or even Oracle Linux): These systems come with caveat emptor by default. Installing such a system prompts you with the message that it comes without warranty (not even the implied warranty of fitness to a particular purpose...), or at least should prompt with such a message.

For type 1 systems we should say that we do not provide "personal, professional service", and for type 2 systems we should remind the users that these systems are caveat emptor by default.

A first attempt:

U&L is a community of users of UNIX and UNIX-like systems. As a community we have no affiliations with vendors of UNIX systems and therefore we do not provide professional support service. All answers are presented in the hope they will be useful but WITHOUT ANY WARRANTY.

Although just adding WITHOUT ANY WARRANTY should be enough, I'd still lean towards we do not provide professional support service. Imagine, for example, that someone using AIX gets an answer here at U&L and the user that gave the answer has an IBM logo as his profile picture.

In theory the user is liable for trademark infringement, but is he liable for damages resulting from his answer?

  • 3
    Commercial software EULAs usually disclaim warranty too (within the limits allowed by whichever laws apply), there's not much difference there between proprietary and free software. As far as support goes, some Linux distributions do come with a support contract. Nitpick: NetBSD has nothing to do with Novell ;-). – Stephen Kitt Aug 9 '16 at 8:38
  • NetBSD is ran by the NetBSD Foundation, a non-profit, and yeah, Red Hat will likely be happy to sell support for their system. But does this distinction matter for SE? If the developers give any warranty or accept responsibility for their system is surely distinct from whether some third party, like a user of SE does for their advice? – ilkkachu Aug 9 '16 at 13:33
  • Thanks guys for the NetBSD, I never used it so I'm clueless on that. Yet, I extended the discussion with a trademark infringement possibility scenario. Contrary to typical linux/*BSD distros, places like IBM or Oracle would not hesitate to throw their army of lawyers at such a user just to make sure he gets the blame. (Mostly to ensure that they do not get the blame) – grochmal Aug 10 '16 at 2:35
  • 1
    I don't understand your distinction between 1 and 2, at least not worded in these terms. Oracle, HP, RedHat, Suse etc. have enterprise supported Linux-based systems, you can get free no-warranty Solaris. Not to mention Apple's BSD based systems and supported BSD or Linux based systems bundled with hardware. – Stéphane Chazelas Aug 10 '16 at 20:39
  • @StéphaneChazelas - I'm mostly concerned with IBM (AIX) and HP in point 1. Their systems are completely theirs and their EULAs are very different from what most *nix users are used to. For example, IBM and HP-UX aims to sell their UNIX systems to customers that want warranty and responsibility for the system from IBM (mostly banks and oil companies). Their kernels are different, not based on Linux or BSD 4.4 in any way (based on earlier BSDs to some extent though). – grochmal Aug 10 '16 at 21:28
0

I think certain cases justify a disclaimer, but it's merely a way for the answerer to advise carefulness in the situation.

The terms of the site protect it from user's mistakes anyway, and abusing disclaimers would probably become annoying.

  • 1
    Yes, the terms of the site protects it, but they don't protect us. In fact, they explicitly say that the responsibility for the content lies with the user who posted it. So if you post an answer with an undetected bug, or with a weird side-effect, the responsibility is yours. – Kusalananda Aug 11 '16 at 10:50
  • 3
    That is actually a bad idea to responsibilize someone considering that even the biggest corporations always refuse any kind of responsibility. Everyone can make mistakes or be misunderstood. – Julie Pelletier Aug 11 '16 at 16:09
0

I think this would be a question for the law stack exchange. According to my sense of justice, no one who gives free advice in a setting like Stackexchange is in any way accountable for damages which occur from people using that knowledge, but courts all over the world may disagree.

If it's possible for users to be held accountable after all (in a legal manner, beyond downvoting ;-) ) then I think all stackexchange sites should include waivers into their terms of use to also protect the actual content authors and not just themselves.

That said, I think it's a highly unlikely scenario that someone would try to sue over a destructive answer.

On the topic of disclaimers: Well, if any, I guess they should be voted on on a per question answer. Because if the question is about something critical (like deleting files) then all answers will be prone to destructive behaviour if they contain any mistakes.

But I think it would be very hard to decide, probably rather subjective, and so probably not productive.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .