Should communities have an “unaddressable mode”?

I’d like to see an “unaddressable mode” for communities: choose a duration, receive a rotating pseudonym, and turn off DMs, mentions, profile discovery, search indexing, and long-term visible history. The goal isn’t disappearing from moderation; it’s making it harder for ordinary members to assemble a durable social graph around someone who needs a break from harassment, social pressure, or compulsive availability.

The platform could retain a stable internal account key, reports, timestamps, and limited case data for moderators, with documented access and retention rules. Implementation gets difficult quickly: how do you prevent abuse without making pseudonym periods trivially linkable, and what happens across federation? ActivityPub supports multiple actors and deletion activities, but remote servers may retain transmitted data. “Deleted” needs to mean deleted from systems you control, not magically erased everywhere.

Would this reduce harm, or mostly create new loopholes? Please disagree, suggest safeguards, or point to a service that already offers something similar.

Federation is the sharp edge. A local mode can reject inbox deliveries, but once a post has crossed to another server, you can only send a deletion request and hope the recipient honors it. I’d separate “unaddressable here” from any promise of universal disappearance.

The social benefit seems plausible, but the interface wording matters. “No mentions” could mean reject addressing, suppress notifications, or hide the mention after delivery. Those produce very different expectations. I’d make the mode visibly time-limited and show users exactly what remains public.

A rotating name is not automatically unlinkability. Rate limits, device signals, writing patterns, and moderator linkage can reconnect periods. That may be acceptable, but the threat model should say whether this protects against casual browsers, stalkers, or the platform itself. One feature cannot promise all three.

I’d want evidence before claiming it reduces harassment or compulsive availability. The design is a reasonable hypothesis, not a demonstrated outcome. A pilot should measure unwanted contact and return rates while also tracking false reports, evasion, and whether people understand the deletion limits.

Discord already has pieces: disable DMs, filtering, blocking, and ignoring. What’s missing is the temporary identity and history boundary. I’m less worried about the button than the support queue when someone asks why their old contacts can still recognize them.