Separate out "edit own messages" and "delete own messages" permission in community permission settings

(#3454) Feature Open: Under consideration community moderation

Thread

Comment by @fluxerhost
RexSystem 3 votes originally by @fluxerhost on GitHub OP 12 replies

Example use case for this feature: running fluxer in a corporate setting, or for a law firm.

"Legal discovery" and evidence preservation for all messages must be preserved. If an employee leaves or is fired, they don't get to take their messages with them. If an employee is involved in a corporate investigation (e.g. harassment), the investigations team should be able to extract all messages sent by the user. Message immutability via expanded ACLs is a good first step to support such safety. It is more technically complex, but also a good idea, to preserve all edit and delete history, however it is also a good feature and easy to implement to add message immutability/making the ACLs around messages more granular.
Comment by @Shardion
RexSystem 3 votes originally by @Shardion on GitHub
I believe that this is a feature which should not be considered on its technical merits alone. It's true that there is probably not much obstructing the implementation, and that it could theoretically be useful to someone for a legitimate purpose. If you look at it like that, of course it'd be a good thing, but technology does not exist in isolation of social realities, and it is easy to misuse it, accidentally or intentionally. This particular feature has a few, listed use-cases, most of which seem to be legitimate, but also a lot of ways in which it can be misused...
  • Create an environment that encourages harassment (the "stand by what you say" idea)
  • Prevent users from burying things, or otherwise leaving them in the past
  • Make it more difficult for users to leave a community, by forcing them to take all or none of their messages with them
  • Admins enable because they, somehow, like it better that way
And, the original concern, of making the platform significantly less usable for a subset of people, still stands. I don't think any of the provided remediations for these issues are effective.
  • A scare screen would be in opposition to regular use for a channel. Either the scare screen is reduced to the point of ineffectiveness, or it is sufficiently annoying to obstruct general discussion. Given the issue of alarm fatigue, I don't think there can be a good middle-ground in designing a scare screen like this, but there hasn't been any user testing, so that might not be true.
    • An alternate design might, alongside a scare screen, have a little notice in the bottom right, like the slowmode indicator. If you do want to do user testing, try adding that...!
  • Edit slowmode either doesn't do anything to stop edit trolling, if you can edit once after sending a message, or, if you can't, obstructs significant real editing usage, like typo corrections and adding context. It also doesn't do anything to solve edit trolling if the slowmode duration is too short, since you can just wait, say, one minute, then edit the message like normal.
  • Edit windows are probably the best idea here, but there are still valid reasons to edit a message, say, five minutes after it's been sent. It might be a low-traffic channel with long messages, like many of the topic channels in Fluxer Labs, or it might be a living message, pinned or linked to, which would benefit from being updated long after its initial sending.
  • Making it exclusive to self-hosted instances has been discussed at length, and I won't bother to repeat it here. Everything that Fluxer (software) can do becomes a part of Fluxer (platform), as soon as federation is involved, so I don't think this is sufficient to stop misuse, and Fluxer Platform AB is at least partially responsible for what features they allow on Fluxer (platform). If Fluxer.com were to choose not to enable this feature, that means there is a significant reason behind the choice, as the default choice is lower-friction, and gives broader access to more features of Fluxer (software). Nobody has posed objections which are not related to safety or data privacy, so it is rather safe to assume that those are the reasons.
  • Sure, you can choose to leave communities and instances that enable this, but that means you're being excluded from them. It would be best to not hand out access to levers that exclude people. Some people would pull it for fun, others would pull it out of malice, but you know that someone is definitely going to pull it.
I also do not accept that the ability to disable message deletion is OK, purely because you can choose to leave a community and delete all of your messages. That adds a lot of friction to what should be a simple action of deleting a single message, which ultimately requires a choice by the user: assuming they can even rejoin the community, do they value getting rid of that one message more than all of the messages they've sent there? You might be surprised by how little attachment users have to their messages, and I, personally, would probably choose to get rid of everything. If anyone chooses to do that, the purpose of the permission has been defeated, and you lost context in previous conversations, and records of real participation in your community. I ultimately believe that these features go against one of Fluxer's goals, to enable easier data privacy for all users. It should be as easy as possible to control exactly what data you have on Fluxer, who can see that data, and what the data they see is. The ability to disable editing or deletion in a channel would, completely unnecessarily, make it more difficult, in service of use-cases that are either already covered by existing solutions, or already obstructed by non-optional features of Fluxer. There is also the potential (the only thing there can be, since the feature doesn't exist yet) for harm, and a significant impact to usability. I have raised the alternative solution, of a Fluxer fork, "Business Fluxer", which adds the feature, and removes support for federation with standard Fluxer, in a way that is not trivially reversible. Ideally, it would be maintained by businesses that require it, or even officially, by Fluxer Platform AB. This fork would already be necessary to use Fluxer in a business context, as Fluxer is making a point of having data privacy features, like the ability to delete your account and all messages that it has ever sent, and these features are always enabled, for all users, on all instances. This would remain true, and unaffected by the presence or absence of editing and deletion permissions, unless Fluxer goes back on this, and makes those features optional, which I believe would significantly ease abuse on a federated network. I don't think the good outweighs the bad here, and being reductive about the bad won't help. You do not have my support.
Comment by @Shardion
RexSystem 2 votes originally by @Shardion on GitHub 2 replies
There is nothing that can enforce an instance to only be able to federate with other instances running the same code, and it doesn't happen automatically, even when incompatible changes are introduced. This is actually why I suggested a fork! If such a fork existed, say, "Business Fluxer", and was made intentionally incompatible, maybe by ripping out the federation code entirely, then the legal use-case could have its own Fluxer, without the data privacy features for compliance reasons, and personal users on the federated network get to keep their data privacy in the safest way possible.
Comment by sodiboo
sodiboo 2 votes edited 1 reply
I think the idea of having a distinct "Business Fluxer" for legal compliance makes sense. It wouldn't really need (at least, technically) to be a separate fork if we just want this one feature. It could be an instance-wide toggle at setup.
Should this instance be immutable for legal compliance reasons? Federation will be disabled, deletion of messages will be disabled (instead, perhaps, they can be "hidden"), edit histories will always be visible to everyone, and channels and guilds can only be archived; not deleted outright
^ or this could be only settable with an environment variable, to prevent encouraging non-bureaucratic users from using it for evil reasons (i'm generally of the opinion that this feature is overall harmful, but i also recognize that it may be Legally Required in certain cases), and also just to prevent asking confusing questions with major irreversible implications to users who don't need it. But it's worth considering if any current Fluxer users even need this feature? Surely none of the commenters here are considering hosting it for business or government use. And, i believe those use cases need specific certification or audits? Even if Fluxer builds the perfect software for them, it doesn't mean Fluxer can be used for them automatically. so, the "Business Fluxer" fork would be the most suitable way forward here, in my opinion. Not for technical reasons (the feature would not be infeasible to implement in upstream), but purely for the compliance reasons of How Such A Product Must Be Developed In The First Place. let someone else handle the compliance side of things. Fluxer HQ can focus on building a chat platform for normal people. or, at least, wait until a business user comes along and wants to help develop the feature for upstream fluxer (in a way that is actually compliant with their policies)
Comment by astromahdi
I think an instance-wide configuration option only on initial setup makes sense. It defeats the purpose of auditing if a non-audited fluxer instance “becomes” an audited instance (the non-audited history “taints” its absolute auditability) Maybe the option can be turned off if it had been enabled during setup -- essentially, a feature that "goes" only one way. I am aware of a user who wants to host a fluxer instance for business use, and is willing to fund/support the development needed for the usecase.
Comment by @Shardion
RexSystem 1 vote originally by @Shardion on GitHub
This use-case probably interacts badly with Fluxer's existing data privacy features, so those will have to be considered as well, when determining if Fluxer should support it or not. I would personally wish for this to be out-of-scope, since I don't want edit- and deletion-blocking to exist in the federated network, but it might be a good basis for an (incompatible!) fork?
Comment by @fluxerhost
RexSystem 1 vote originally by @fluxerhost on GitHub OP
I don't think forking is the right suggestion for harmless features, especially expanding existing ACLs into granular ACLs. I don't think a corporate entity would federate with internet-discoverable instances anyway (probably only with other corporate instances), and expanded ACLs do not impact federation either way for non-corporate use caes. In the case of a corporate use case, "leave and delete my messages" from a community would probably have to be toggle-able at the instance level (new gh discussion), and even deleting an account would have to be toggle-able at the instance level as well (new gh discussion). Those, however, are separate feature adjustment discussions, i think -- this github discussion is focused on a portion of the features needed to support such a use case: expanding the ACL granularity for message actions (to split out edit and delete permissions).
Comment by @Shardion
RexSystem 1 vote originally by @Shardion on GitHub
We are discussing the potential harm of this feature right now. These features probably do not impact federation on a technical level, but they do impact the resulting federated Fluxer platform, since people can access communities within an instance through federation, being subject to its own configuration, as opposed to that of their home instance. I feel that it is against the goals of Fluxer to allow instances which disable data privacy features to exist in the federated network. It's important to note that, in federation, all features will be implemented eventually. It can't really be helped, so federation must be designed in a way to minimize the impact of instances which operate in an undesired or unintended way, and, to minimize the chances that an instance operator will attempt to do that. This is why I feel that having it as a toggle which Fluxer.com disables is not sufficient for this feature to exist. Not only would it require extra development work on Fluxer Platform AB's side, for a feature that they don't intend to use, and doesn't further their goals, but it also significantly lowers the friction involved to have this feature, from "fork Fluxer and add it yourself", to "go into the container and delete the line preventing federation without data privacy".
Comment by @fluxerhost
RexSystem 1 vote originally by @fluxerhost on GitHub OP
I don't think this feature has any harm. Expanded ACLs are always better than opaque ACLs. This feature does not impact privacy either. Channel visibility is still controlled by admins and roles, via ACLs. Perhaps only instances running the same code will be able to federate with each other - which is why it is important that this feature is treated seriously and part of the official codebase, even if unused on the main fluxer.com instance. I think it is sufficient for this feature to exist. The use case is a real one.
Comment by @fluxerhost
RexSystem 1 vote originally by @fluxerhost on GitHub OP
I think suggesting a fork is not the right move for anyone involved and that a fork for expanded ACLs a waste of effort. I think this is a valid feature to exist in the main Fluxer codebase. Again, this feature does not impact privacy or safety negatively.
Comment by @fluxerhost
RexSystem 1 vote originally by @fluxerhost on GitHub OP
To clarify about the potential privacy and "harm" concerns: The concerns listed about "harm" are both theoretical and potential, not actual or clearly listed. Any administrator can use any feature in a harmful way - such is the way of any technology. How a person uses any technology is not in the realm of consideration of a technical feature such as granular ACLs. The concerns listed about privacy are also not relevant to this feature - not only would the feature be toggle-able at the instance level, and even further, would be toggle-able at the community, and ultimately at the ACL level, as proposed here, per role or channel (like other permissions), but also, the "right to be forgotten" is covered by two features already: the community-leave-and-delete-all-messages feature, which would be unaltered by this feature, or the delete-my-account-and-all-data feature, which would also be unaltered by this specific feature request, as proposed. I think vkyfox explained about safety and privacy concerns well as well: https://github.com/orgs/fluxerapp/discussions/3200#discussioncomment-18748231
Comment by @fluxerhost
RexSystem 1 vote originally by @fluxerhost on GitHub OP
The good outweighs the bad here, because there is no "bad" for granularizing ACLs on existing actions.
I believe that this is a feature which should not be considered on its technical merits alone.
I, on the other hand, believe that this is a feature that can easily be considered on technical merits alone, because it is simply granularizing existing ACLs and actions.
It's true that there is probably not much obstructing the implementation, and that it could theoretically be useful to someone for a legitimate purpose. If you look at it like that, of course it'd be a good thing,
Well, that much is clear.
but technology does not exist in isolation of social realities, and it is easy to misuse it, accidentally or intentionally.
Misuse is the fault, concern, and problem of the misuser, not the developer, of a technology. Computers can be misused. Should we stop manufacturing computers? I'm not going to elevate this to a theoretical discussion, because that both misses the point of a technical feature request and is a waste of time. I shall now show you how an existing feature can be "misused" by a "bad" administrator: channel visibility. This is a feature that already exists in the Fluxer codebase, ACLs, etc. I have seen this happen in a community already:
  1. Users chat in #chat. They send messages.
  2. Administrator creates #chat-2
  3. Administrator removes channel read/visibility access for @everyone for #chat
  4. Users' messages are now "locked away" in #chat, unable to be deleted, essentially immutable. The only way for a user to delete those messages is leaving the community and deleting all their messages, or, deleting their account and deleting all their messages, as I proposed above.
This is already possible with existing ACLs. The feature proposed in this discussion post does not introduce any new "risks" of any kind. Anyone can use or "misuse" any feature that already exists for any reason they desire. It is up to users to participate in communities that they trust and are comfortable with the administration policies of. Furthermore, "living messages" are a "misuse of necessity" of a chat platform. Just because all one has is a hammer, doesn't mean everything should be treated as a nail. Will respond more tomorrow.
Comment by @fluxerhost
RexSystem 1 vote originally by @fluxerhost on GitHub OP
This particular feature has a few, listed use-cases, most of which seem to be legitimate, but also a lot of ways in which it can be misused...
Again, misuse is in the hands of, and the problem of, the misuser, not the developer, of a technology. Computers should not be banned because people find ways to misuse them.
Create an environment that encourages harassment (the "stand by what you say" idea)
This is a mischaracterization of the proposal and furthermore a social problem, not a technical one. A developer can not control the users of any feature.
Prevent users from burying things, or otherwise leaving them in the past
This is not the problem of the developer of a technology, or the platform itself.
Make it more difficult for users to leave a community, by forcing them to take all or none of their messages with them
I don't see how this is the case. The "leave a community and delete all messages" feature is untouched, as well as the "delete my account and delete all messages/content" feature is also untouched in this proposal.
Admins enable because they, somehow, like it better that way
Is this not within the realm of what makes an administrator of a community, an administrator? It's their community after all.
And, the original concern, of making the platform significantly less usable for a subset of people, still stands.
I don't see how this makes anything less usable.
I don't think any of the provided remediations for these issues are effective.
I don't think those are issues that are the concern of the developer or the platform.
A scare screen would be in opposition to regular use for a channel. Either the scare screen is reduced to the point of ineffectiveness, or it is sufficiently annoying to obstruct general discussion.
Such a screen already exists for NSFW channels. This is not a point.
Given the issue of alarm fatigue, I don't think there can be a good middle-ground in designing a scare screen like this, but there hasn't been any user testing, so that might not be true.
No need to invoke "alarm fatigue" here. This is also a feature request, not a PR, so there is no building of code ahead of time or testing. That is up to the developer/team.
An alternate design might, alongside a scare screen, have a little notice in the bottom right, like the slowmode indicator. If you do want to do user testing, try adding that...!
Thank you for the suggestion. Perhaps the development team can take this into account as an option/possibility.
Edit slowmode either doesn't do anything to stop edit trolling, if you can edit once after sending a message, or, if you can't, obstructs significant real editing usage, like typo corrections and adding context. It also doesn't do anything to solve edit trolling if the slowmode duration is too short, since you can just wait, say, one minute, then edit the message like normal.
Agreed, edit slowmode isn't a good solution to either the symptom ("edit trolling" as mentioned in OP) or the root cause.
Edit windows are probably the best idea here, but there are still valid reasons to edit a message, say, five minutes after it's been sent. It might be a low-traffic channel with long messages, like many of the topic channels in Fluxer Labs, or it might be a living message, pinned or linked to, which would benefit from being updated long after its initial sending.
Perhaps messages in a chat application are being misused as is ("living messages") out of necessity and feature scarcity. Perhaps pinned posts would be able to be edited as another toggle/ACL, only while they are pinned.
Making it exclusive to self-hosted instances has been discussed at length, and I won't bother to repeat it here. Everything that Fluxer (software) can do becomes a part of Fluxer (platform), as soon as federation is involved, so I don't think this is sufficient to stop misuse, and Fluxer Platform AB is at least partially responsible for what features they allow on Fluxer (platform).
"Misuse" of existing features has already been covered, so I won't bother to repeat it here. Misuse is not the problem of the developer of the software.
If Fluxer.com were to choose not to enable this feature, that means there is a significant reason behind the choice, as the default choice is lower-friction, and gives broader access to more features of Fluxer (software). Nobody has posed objections which are not related to safety or data privacy, so it is rather safe to assume that those are the reasons.
No, the objections people have posed "related" to safety or data privacy are, seemingly, grasping for the terms safety or data privacy without substance, as what appears to be a knee-jerk reaction to a feature they don't seem to agree with. Fluxer.com, additionally, could choose to enable or disable any subset of features on its instance for any reason, such as: preference. Not everything declined is dangerous.
Sure, you can choose to leave communities and instances that enable this, but that means you're being excluded from them.
And that's the choice of the user. Nobody is obligated to participate in communities they don't like. Some people don't like bots. Should bots be banned on Fluxer?
It would be best to not hand out access to levers that exclude people. Some people would pull it for fun, others would pull it out of malice, but you know that someone is definitely going to pull it.
Such levers already exist in the software. This is no different.
I also do not accept that the ability to disable message deletion is OK, purely because you can choose to leave a community and delete all of your messages.
You do not have to accept it. Not every feature will be appealing to every user.
That adds a lot of friction to what should be a simple action of deleting a single message, which ultimately requires a choice by the user: assuming they can even rejoin the community, do they value getting rid of that one message more than all of the messages they've sent there? You might be surprised by how little attachment users have to their messages, and I, personally, would probably choose to get rid of everything.
Well, then, okay - feel free to get rid of everything, that's a valid choice.
If anyone chooses to do that, the purpose of the permission has been defeated, and you lost context in previous conversations, and records of real participation in your community.
That is fine, and up to the user, to delete their messages upon leaving a community or deleting their account.
I ultimately believe that these features go against one of Fluxer's goals, to enable easier data privacy for all users.
I ultimately believe that this feature goes for one of Fluxer's goals, and that's to empower its users.
It should be as easy as possible to control exactly what data you have on Fluxer, who can see that data, and what the data they see is.
"should" - perhaps you have feature requests you would like to submit as well? Otherwise, I don't think either of us is the one in charge of determining (or being the final judge of) either the goals of Fluxer or "should"s.
The ability to disable editing or deletion in a channel would, completely unnecessarily, make it more difficult, in service of use-cases that are either already covered by existing solutions, or already obstructed by non-optional features of Fluxer. There is also the potential (the only thing there can be, since the feature doesn't exist yet) for harm, and a significant impact to usability.
I still disagree with the arguments of "potential" and unrealized "harm". It seems like fearmongering against a suggestion that is not desired, for whatever reason.
I have raised the alternative solution, of a Fluxer fork, "Business Fluxer", which adds the feature, and removes support for federation with standard Fluxer, in a way that is not trivially reversible.
This sounds like unnecessary fragmentation and effort for the development team. Also outside the scope of this discussion.
Ideally, it would be maintained by businesses that require it,
No one will, this idea is a non-starter. Also outside the scope of this discussion.
or even officially, by Fluxer Platform AB.
This may be possible, theoretically, but practically, unfortunately not. At least not until many more resources are under the command of Fluxer Platform AB, in my opinion. Unnecessary product segmentation. Also outside the scope of this discussion.
This fork would already be necessary to use Fluxer in a business context, as Fluxer is making a point of having data privacy features, like the ability to delete your account and all messages that it has ever sent, and these features are always enabled, for all users, on all instances.
These features could be placed behind toggles in the admin panel so that a business use case is possible (which is a valid use case for communications software such as Fluxer). That, however, is outside the scope of this discussion around expanded message ACLs, so I will not explore it here.
This would remain true, and unaffected by the presence or absence of editing and deletion permissions, unless Fluxer goes back on this, and makes those features optional, which I believe would significantly ease abuse on a federated network.
Again, abuse is in the hands of and fault of the people taking those actions, not the developers of the tool or platform.
I don't think the good outweighs the bad here, and being reductive about the bad won't help. You do not have my support.
I think the good outweighs the bad, because there is no "bad" to expanding an existing opaque ACL into granular permissions. As for not having your support on this: such is natural. Nobody agrees 100% on every topic. Monocultures die after all. Thank you for your feedback, however.