did-method-plc
PLC Operational Updates (October 2025)
Cross-posting from: https://github.com/bluesky-social/atproto/discussions/4316
Our near-term roadmap for PLC involves implementing a fully-validating read-replica service.
It’s important that replicas agree on which PLC operations are valid and which are not. Due to past bugs in the reference implementation, the live plc.directory database has accumulated a number of invalid operations.
We will be removing these invalid operations at some point after 2025-11-06 (i.e. at least 7 days from now). See here for technical details and listings of the invalid operations. We believe that these operations were all created by developers experimenting with the PLC mechanism, and do not anticipate that anyone will be adversely affected by their removal. Upon their removal, we'll make an archive available with all the operations removed in this event. (Update: an archive is available here: https://archive.org/details/removed_plc_ops_2025-11-14.log )
Some of the invalid operations are updates to otherwise-valid DIDs. In these cases, the DID will be reverted to its last valid state.
Additionally, clarifications have been made to the specification to align it with the reference implementation, minimizing ambiguities regarding operation validity. The latest version of the specification is available at https://web.plc.directory/spec/v0.1/did-plc [^1]
Planned Future Changes
The /export API will be augmented to allow for low-latency synchronization with replica instances (likely using websockets).
The PLC directory might assign a monotonic sequence number to each accepted operation, as part of a future implementation of Transparency Logging.
P.S.
We're aware of a number of community efforts towards replicas/mirrors of PLC, such as https://plc.wtf/ and https://plcbundle.atscan.net/ . Awesome! It's up to them how they decide to handle the aforementioned invalid operations.
[^1]: Yes, the version number in the URL is wrong! This will be corrected eventually.
Thought I'd drop a link to a thread I started on the community discourse, where those of us running PLC mirrors are sharing our experience, insights, and filters to drop the abuse operations against the PLC. This certainly goes well beyond what this issue is about, and we also want to bring the results of this issue back to that thread.
https://discourse.atprotocol.community/t/plc-filtering-for-mirrors-guidelines-rulesets-and-experiences/241
