The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.
Let's close this and get this over with now.
There is consensus that administrators can block for a pattern of local abuse which includes vandalism.
There is consensus that administrators can block open proxies if they have been used on Wikidata.
There is no consensus regarding administrators blocking per a consensus with strong opposition to a one week requirement. Administrators may block, mentioning a consensus if relevant and would back up the initial blocking rationale although this does not remove the fact blocks should be done per policy.
Global sysops and stewards may use their block tool to block only per cross-wiki abuse if necessary.
Said cross-wiki blocks must be as a result of the user having a likelihood of hitting Wikidata or already having done so.
Administrators may block for edit warring which is defined as repeated reverts over a short period of time. As Wikidata's editing system is different, reverts are defined as undoing a series of edits or an idea put forward the user. This may results in multiple reverts for a single idea as such thus would not constitute and edit warring. If content will be oversighted after, edit warring does not apply. Administrators may fully protect an item or property if they seem necessary and it would force the users in question to discuss on the talk page as blocking would prevent constructive discussion. Administrators should apply discretion when dealing with a case of edit warring as strict numerical rules are not always accurate when dealing with Wikidata.
Administrators may block for harassment and personal attacks.
Oversighters are authorized to block users per oversighted content while global sysops and stewards are not.
Administrators may block a user when required when no community consensus has been sought or when no policy properly governs the block assuming the administrator seeks feedback at a public venue and can fully explain why they felt the block was necessary to prevent disruption.
Administrators may revoke talk page and email access and they feel necessary if the use is not constructive or carries on the same behaviour the user was original blocked for.
Administrators may block unauthorized and malfunctioning bots as they feel necessary. John F. Lewis (talk) 00:18, 27 September 2014 (UTC)
Lately, Wikidata:Use common sense has not been adequate to prevent discord in our community. No specific examples will be named in order to avoid implying that these disputes are ongoing. But just for example, our blocking policy has not been heeded, as it is only at the level of a guideline; besides blocking for sockpuppetry, which was ratified per Wikidata:Alternate accounts, most blocks beyond vandalism and spam, often even including blocks of bots, become problematic.
Another problem is our lack of definition of what exactly edit warring constitutes, because a "revert" is ill-defined with items and properties. Sometimes one series of edits may be done by a "Restored revision..." edit, which may or may not be considered the same as simply reverting them with multiple "Undid revision..." edits.
These problems are due to the multicultural nature of our wiki, because different home wikis have different conventions regarding user conduct policies. Therefore, it has become time to attempt to codify these.
This RfC is in the form of both any existing content of the corresponding guideline pages and new amendments. Existing guidelines are prefixed with "Existing:". Whatever achieves consensus will form the new policy.
This RfC is to stay open for at least four weeks. The first phase will last 2 weeks and will be an opportunity to add new options as necessary. After this weeks, no more new options may be added and in the second phase discussion must focus on passing or rejecting the existing options. All options should be translated with the translate extension before the end of the first phase.
For the purposes of the above section, local abuse would be defined in accordance with this section. The options are not mutually exclusive. All of the following options that pass will be included in the definition.
The extensions can be switched on and off per community decision. Torblock extension has settings that allow already registered users to edit using ToR. It's up to community to decide, it is not a technical barrier. -- Vlsergey (talk) 08:34, 24 July 2014 (UTC)
@Vlsergey: If you wish for that, then please add a separate proposal for that. Opposing this will just lead to hindrance of necessary abuse prevention, but I do agree that admins should be responsive when handing out IP block exemption. My opinion is that they are, as requests at WD:RFP are handled pretty quickly, but if not, then this is out of the scope of this RfC and needs to be addressed in an amendment to Wikidata:IP block exemption, not hindrance of blocking of open proxies.--Jasper Deng (talk) 08:37, 24 July 2014 (UTC)
I'm opposing to fully implement this rule and support to implement it only partially -- to allow already registered users to use open proxies, but prohibit non-registered from using them for edit or for creation of new accounts. From my point of view, it is enough to fight vandalism and prevent abuse. Well, at least until someone will show that this is not enough. -- Vlsergey (talk) 22:13, 24 July 2014 (UTC)
@Jasper Deng: For me this is about removing the need for Wikidata:IP block exemption permission (at least for Tor), so this would not be an ammendment to that permission. My concern is not responsivenes of admins. I would be fine with splitting this up into "open proxies except Tor" and "no block exception needed with Tor for logged in users", but I don't feel confident to do this now myself. Which part of Wikidata policy would support the current setting of mw:Extension:Torblock regarding needing block exception to edit over Tor if not this part that is currently discussed here? JanZerebecki (talk) 16:46, 31 August 2014 (UTC)
@JanZerebecki: Whenever torblock is enabled you must have its exemption permission to edit through it; if that permission were automatically assigned then the extension is effectively useless. I want torblock to be a separate discussion.
The reason why Tor is blocked on almost every Wikimedia wiki is because of the abuse it's caused for us.
With that said you're more than welcome to re-amend this after this RfC with... another RfC.--Jasper Deng (talk) 17:12, 31 August 2014 (UTC)
@Jasper Deng: I'm fine with anything here related to Tor being a separate discussion. Which would require amending the description of this "Open proxies" policy item with "This explicitly excludes any decision regarding access through Tor e.g. either for or against blocking and/or necessity of block exceptions or not.". Regarding your Tor related arguments: Users via Tor that are not logged in would still be blocked by the extension so it would not be useless. The user accounts used to log in via Tor would either have been created via a non-Tor IP which still can be blocked or via mailing a account request. Obviously such an account can still be blocked. So this would in no way limit the ability to fight abuse. JanZerebecki (talk) 20:21, 13 September 2014 (UTC)
@JanZerebecki: But some people don't want TorBlock to be enabled at all. And even if Torblock were still enabled, removing admins' ability to manually block nodes would still hinder anti-abuse because the extension is not 100% effective. The English Wikipedia has a bot specifically to block Tor nodes not blocked by the extension.--Jasper Deng (talk) 20:27, 13 September 2014 (UTC)
@Jasper Deng: With our current state of abuse fighting infrastructure I think it is fine to block Tor exit nodes that the extension somehow misses, as long as the manual block gets removed as soon as the extension picks that IP up. Although I would prefer the extension being fixed. JanZerebecki (talk) 20:56, 13 September 2014 (UTC)
Conditionally As long as admins are responsive about allowing for IP-block-exempt to legit users. —Justin (koavf)❤T☮C☺M☯ 00:12, 24 July 2014 (UTC)
If there is a legitimate need to use ToR (i.e. you live in China), you can request an IP block exemption from stewards at m:SRGP. --Rschen7754 23:45, 28 July 2014 (UTC)
I’m curious but how is an IP-based exemption going to work for unblocking Tor in 2014? We’ve already officially run out of IPv4 addresses and any user’s IP address can change any time and any single IP could be an entire university.—Al12si (talk) 08:52, 30 July 2014 (UTC)
There should be a valid reason to block the user, not just consensus of the community. The consensus we may seek is there a valid reason to block or not. There should be not any community-driven blocks without valid reason established by rules. -- Vlsergey (talk) 08:04, 24 July 2014 (UTC)
1 week is too long. --Rschen7754 21:42, 24 July 2014 (UTC)
A week is really long. I dislike community blocks, I would expect that users have to know more or less beforehand how to behave without being blocked, while community votes could change the "rules". Lymantria (talk) 16:42, 27 July 2014 (UTC)
Under what circumstances would a community consensus be used solely to block a user, if any? —Wylve (talk) 12:31, 24 July 2014 (UTC)
The community would only do so when it deems a user to be "disruptive", i.e. working against the interests of Wikidata - but not necessarily included in any of the other reasons. That may mean different things in the future; it's an elastic clause, really.--Jasper Deng (talk) 17:34, 24 July 2014 (UTC)
The existing guideline, but without the requirement of one week. Administrators may block any user for which there has been a broad community consensus over any reasonable amount of time, even if the reason is not covered by this policy.
Global sysops and stewards may only block for local and cross-wiki abuse as defined here, and must defer to local administrators for any other reason. This is already specified in Wikidata:Administrators, but not entirely clear about what they can and cannot do.
Administrators may block for cross-wiki abuse if at least one of the following options are chosen. The options below are not mutually exclusive; all that pass will be included in the definition for the purposes of blocking.
Cross-wiki abuse is not blockable on Wikidata unless there is likelihood that the user would hit Wikidata, or the user has already caused local abuse on Wikidata. This specifically excludes cross-wiki sockpuppets that do not have a history of abuse here (i.e. blocks may not be made simply to "autoblock the user globally").
if the user has already caused local abuse - yes. The rest of this proposal is too vague for me even to consider supporting it in whole; that is, what would classify as cross-wiki abuse? --Snaevar (talk) 11:02, 24 July 2014 (UTC)
Strongly at least it's a famous interwiki abuse e.g. called Kagemusha/影武者. --Liuxinyu970226 (talk) 07:23, 25 July 2014 (UTC)
Supporting with an suggestion for different wording. "such an offence" should be replaced with "vandalism as defined in Wikidata:Vandalism" or anything along that line. --Snaevar (talk) 11:02, 24 July 2014 (UTC)
If an administrator finds it necessary to block a user immediately to protect Wikidata, but without (but not against) community consensus nor for any of the other reasons in this policy, the block can be made, but the administrator must post notice of the action at Wikidata:Administrators' noticeboard for review by the community as soon as reasonably possible after the block. IRC and other off-wiki mediums are not sufficient for this purpose.
Any discussion or edits not related to the block or the conduct of the blocked user (specifically excluding the conduct of others) that led to the block is not appropriate use of the user talk page. Misuse also includes repeated frivolous unblock requests or anything that would be otherwise blockable, such as spam. Any uninvolved administrator may, at his or her discretion, disable talk page access for a blocked user misusing his or her talk page.
When it's clear the user is just trying to be disruptive, talk page access can be removed.--Jasper Deng (talk) 08:08, 24 July 2014 (UTC)
Basically per Jasper Deng. For example, an response from an blocked user to an discussion on his chat page is acceptable (even if the subject is not about the block itself), but any disruptive behavior is not.--Snaevar (talk) 11:02, 24 July 2014 (UTC)
don't see the reason to prevent user from using his talk page, unless it's against other rules (i.e. spam, vandalism, self-hosting, etc.) -- Vlsergey (talk) 08:08, 24 July 2014 (UTC)
This clause is incomprehensible, and my translator query has not been responded to after 2 days when the first stage is supposed to be almost over. As a translator I cannot support this.—Al12si (talk) 08:40, 30 July 2014 (UTC)
Administrators may block obviously malfunctioning bots, or bots not authorized by the bot policy. Unless the bot is a vandalism or other bot causing local abuse, the block should be soft (without autoblock) and the bot operator contacted (if known).
Malfunctioning -- yes, unauthorized -- "no". We need to review bot policy and make much lighter version of it (comparing to Wikipedia rules), because Wikidata, from my point of view, are edited and will be edited mostly my bots, not by people. De-facto a lot of bot edits are already here without community consensus. For example, all widar edits are de-facto bot-alike edits -- there are a lot of those, they are all alike, and still, no prior consensus for them. -- Vlsergey (talk) 08:12, 24 July 2014 (UTC)
The actions are clearly done in bad faith and they deface Wikidata pages, including but not limited to the insertion of gibberish, clear libel, or inappropriate obscenities into items, properties, or any other page.
Lymantria (talk) 08:15, 29 July 2014 (UTC) I like the general description better. I like the examples. Bad faith however is something we can at best be assuming and should not be the central part of the definition. We cannot read minds.
per Lymantria -- Bene*talk 14:15, 30 July 2014 (UTC)
What about incorrect facts? Or someone not willing to listen to consensus? --Rschen7754 06:49, 31 July 2014 (UTC)
Lymantria has a point; also, as Snaevar says, gibberish could easily be a test edit. MJ94 (talk) 06:53, 31 July 2014 (UTC)
Edit warring is defined as repeat reverts over a short period of time by two or more users when all involved users knowingly fail to resolve the content dispute by discussion. Reverts would be defined below.
in ruwiki there is much simple definition: "edit war begins when the user returns the edit that was lately previously canceled without prior consensus". No need to define "repeat" or "reverts". -- Vlsergey (talk) 08:14, 24 July 2014 (UTC)
Lymantria (talk) 08:25, 29 July 2014 (UTC) If only simple things as interpunction or capitalization are changed (other pages), it should still count as a revert. Essentially we are better off without this definition.
Basically per Lymantria but I mean the other stuff (items, properties and translations) too. It’s way too hard to track history on items, properties and translations and it’s way too easy to “revert” without even a clue you’ve technically reverted. When you can’t even see all the properties on the screen (and when some items are even displayed incorrectly) you can’t penalize people for “reverts”.—Al12si (talk) 09:01, 30 July 2014 (UTC)
Isn’t this definition backwards? Do I assume “On a property or item, undoing a single series of edits by one user constitutes a revert” is what this is trying to say?—Al12si (talk) 19:04, 26 July 2014 (UTC)
A general rule of thumb is that more than three reverts in a single 24-hour period by one user in the same dispute is grounds for blocking. This does not define edit warring and blocks may be made for three or less reverts, or four or more reverts just outside the 24-hour period.
I don't like idea of 3RR, since (I think) it makes administrative work more robotic even though it will make admin's work more easier. — by Revicomplaint? at 02:32, 28 July 2014 (UTC)
A "general rule of thumb" should not become a policy imho. -- Bene*talk 10:17, 28 July 2014 (UTC)
Lymantria (talk) 08:26, 29 July 2014 (UTC) If a larger number of items is involved than 3RR is a strange rule. What is the same dispute? If I revert a disputed change of Property on a series of 10 items, is it a violation of 3RR?
This assumes a specific approach to adding data to Wikidata. In a previous RfC it was discussed that there is a correlation between the number of edits and the potential of errors. Given that I have at this time 1,230,652 edits and given that on a "good" day there can be as many as 100K more edits, three reverts can be from many moths, weeks and days ago. This rule feels therefore like nothing but "finding a stick to hit a dog".
That RfC was to discuss exactly this situation. There was no comment so it rules fine. At issue is that the error rate of what I do is low. When things get off track it may be spectacularly so for a very limited subset of data. It then helps when we communicate, analyse what went wrong so that remedial action can be taken. However, we seem to have decided that that is not necessary. If anything communicating in an aggressive way is the norm.
This RfC is not an RfC as you can read; you feel that it is needed and consequently are not inclined to discuss or consult. Consequently this is very much a power play and certainly not a request. Thanks, GerardM (talk) 09:50, 25 July 2014 (UTC)
@GerardM: If you don't like "aggressive way"'s of commenting, please retract your last sentence. And I'd appreciate it if you could propose alternatives rather than complaining about existing ones.
I personally think WIDAR needs to be regulated too, but that's for another RfC. To be nice, I will not talk about your particular use of it.--Jasper Deng (talk) 10:00, 25 July 2014 (UTC)
This is totally unclear to me, I rather like common sense as guidance here. Lymantria (talk) 08:33, 29 July 2014 (UTC)
Per Snaevar and Lymantria, please UCS. -- Bene*talk 14:24, 30 July 2014 (UTC)
This Option seems to contain a rule to punish all edit warriers equally, unless there is an evidence that one participant mus be excluded e. g. one opponent can prove a prior consent or has posted on a discussion page prior to the editwar. However this Option must be reworded to make things clear.--Giftzwerg 88 (talk) 17:31, 6 September 2014 (UTC)
Exception for oversightable material, vandalism, and spam
Reverting any material that must be subsequently oversighted, any vandalism, or any spam is specifically excluded from the definition of edit warring.
Users reverting others' edits to their userspace (User:Username and User talk:Username, along with subpages) are exempt from the definition of edit warring, except that they may not revert removals of obviously inappropriate content (particularly libel), nor remove block notices or unblock requests while the block is still in effect.
If an administrator posts a comment on a user talk page specifically asking against removal of the comment, that comment is not subject to the above exemption. Administrators are to exercise this exemption solely to ensure that the user in question is acknowledging and addressing the concerns in question.
@Snaevar: By the way, the RfC is specifically designed so that any objections of that kind can be added as new proposals without ambiguity over whether it got consensus or not.--Jasper Deng (talk) 17:40, 24 July 2014 (UTC)
semi-protection should be preferred when it denies access for all parties involved to the page in question. An full protection should not be done in this case unless 4 or more accounts or ips (disregarding the fact that there could be one individual behind several ips) are involved in the edit war. Full page protections should only be active for a limited time (maybe 2 weeks max).--Snaevar (talk) 11:02, 24 July 2014 (UTC)
Administrators are specifically permitted to and advised to exercise discretion when dealing with edit warring. This includes warning users before any sanctions, and being able to refrain from actions when the edit war has since stopped, or on the other hand, blocking even if the three-revert rule is not violated.
Personal attacks are statements deliberately intended to hurt the feelings of someone else, which does not include good-faith criticism with evidence. Except in severe cases, and particularly when a cultural barrier may have caused a misunderstanding, users making personal attacks should be warned before any actioning.
prefer much simple rule: "don't discuss other editors" (with exception of RfA and admin-flag-granting discussion pages) -- Vlsergey (talk) 08:19, 24 July 2014 (UTC)
Peronal attacks are attacks to the (supposed) person of the user involved, in stead of critisism on his/her edits or actions. I miss that point and I do not think it is a good idea to add POV words like "deliberately" and "good-faith". Lymantria (talk) 08:45, 29 July 2014 (UTC)
Harassment is defined as a pattern of repeated offensive behavior that clearly is targeted against another person or group, in the eyes of any reasonable observer. This may or may not include actual direct personal attacks.