If I did, would you try to understand our vision and our protocol, or would you just desperately try to find ways to discredit our efforts to built user-sovereign tech? We believe that users have a CHOICE where their data is hosted, we don't believe that we need to FORCE users into maximum self-sovereign solutions. But giving an option and Credible Exit at all times is key. So yeah, once you find real flaws, I'd appreciate a well-meant issue report or PR, which is not just AI slop, but actual useful work. I'm not gonna waste any time fighting anybody that has the same mission as I have but follows different means to achieve that goal. That said, the fact that all of you are so passionate about this topic is also a good sign, so let's use that energy and focus to building better protocols and tooling

> secrets in unencrypted MMKV storage This could probably be improved. We're taking it up, thanks for the report. > - Cloudflare binary downloaded at runtime with no checksum? Why? That's a supply chain risk for every Umbrel self-hoster. Very much work in progress, we haven't announced the Umbrel integration anywhere yet and are still in development. For now we recommend not to use that project yet until we release our first Alpha version. > Neo4j shipping with hardcoded password User is encouraged to configure his setup when self-hosting. Don't open up your machine to the internet without understanding what your doing. No big difference to many other systems. A password change could be enforced, there is room for improvement, yes. > - GCS backend with no auditable configuration surface? GCS is an implementation detail of one hosted deployment (Synonym), not a requirement of self-hosting. Our vision is to have many independent operators to run instances on infrastructure they control, using whatever platform fits their needs. We can provide deployment guidance, but operators ultimately control their own configuration. Real self-sovereignty comes from the ability to self-host or choose a host you trust, not from visibility into any particular operator’s production config. We may explore additional transparency and deployment documentation in the future. > - you said users have a choice where their data is hosted. True for the homeserver. But is there a credible alternative to Synonym's Nexus for discoverability? Without it, content exists but nobody can find it. Self-hosting a homeserver without an accessible Nexus isn't a real credible exit for most users. Discoverability does require indexing. A homeserver can host your data, but an app still needs some way to aggregate and process the data. Simple apps can often do that on the fly; For more complex apps it's a good idea to externalize that work into an asynchronous indexer, otherwise the user experience becomes slow or incomplete. That aggregator / indexer can be run anywhere. It could be self-hosted (even though that's not yet as trivial as we'd like it to be). For pubky.app that aggregator/indexer is called Nexus. any users may choose a managed Nexus because it is convenient, while others may prefer to run or use an alternative operator’s Nexus. Many people run their own Bitcoin node, but for indexing they still rely on 3rd-party-hosted block explorers, even though they have all the raw data locally, but they don't want to run the indexer themselves. Offering indexing for self-hosting as well as providing a well-run default seems like a fair deal to users. Do you think it should be done differently? > These aren't philosophical objections to the vision of 'your' protocol. Exactly, you pointed out implementation details in young projects that are still under active development, so perfect stability is not expected. What is your view on the broader picture, the fundamental protocol and vision?

As for the neo4j password, would you share with me in which project you found that config please? I assume it came from pubky-docker, which is intedended to be used as dev tooling only, so that hardcoded password makes sense. Neither you nor your LLM understood that this is just dev tooling it seems, so I'm making it more explicit to avoid future confusion:

GitHub

Add development-only warning to README by gcomte · Pull Request #27 · pubky/pubky-docker

Summary Add a prominent README warning that this project is intended for local development and experimentation only Clarify that it should not be ...

Thank you for moving Pubky forward with us.

As for the neo4j password, would you share with me in which project you found that config please? I assume it came from pubky-docker, which is intedended to be used as dev tooling only, so that hardcoded password makes sense. Neither you nor your LLM understood that this is just dev tooling it seems, so I'm making it more explicit to avoid future confusion:

GitHub

Add development-only warning to README by gcomte · Pull Request #27 · pubky/pubky-docker

Summary Add a prominent README warning that this project is intended for local development and experimentation only Clarify that it should not be ...

Thank you for moving Pubky forward with us.

My concern stands but scoped correctly: it's that pubky-docker itself bills as "one click deployments" in its README, which invites non-experts to run dev tooling directly.