News for hackers. Stuff that matters.
Re:Show HN: Permify – Open-source authorization service base... by jzelinskie on Thursday July 14, 2022 @04:01PM (#32100821)
SpiceDB does fully support all consistency described in the Zanzibar paper[0] and even allows the requests to specify consistency on the fly[1]. We've designed around this from the start because it'd be very difficult to add after the fact. We also built a CI pipeline that leverages ChaosMesh[2] to test for the New Enemy Problem.

[0]: https://docs.authzed.com/reference/zedtokens-and-zookies

[1]: https://docs.authzed.com/reference/api-consistency

[2]: https://chaos-mesh.org

Re:Show HN: Permify – Open-source authorization service base... by dastbe on Thursday July 14, 2022 @04:29PM (#32101054)
imo i would add to your docs more details around what to do when you ex. update a relationship but fail to persist a zedtoken, or at least the options and what the ramifications are. these are things that people may not truly reason through and your team has the most context on how these decisions can affect user experience.
Re:Show HN: Permify – Open-source authorization service base... by jzelinskie on Friday July 15, 2022 @02:14AM (#32105143)
Thanks. This is great feedback. We've written a blog post[0] about some of strategies developers are using to write relationships, but we haven't covered explicitly the failure scenarios.

For the scenario that you've described, you should be able to retry persisting the zedtoken, but if there's some fundamental reason that's not working, you can always perform a fully consistency check and persist the resulting value -- fresher zedtokens are always safe.

[0]: https://authzed.com/blog/writing-relationships-to-spicedb/

Re:Show HN: Permify – Open-source authorization service base... by jchw on Thursday July 14, 2022 @04:29PM (#32101059)
I hate to detract from the Show HN post, but frankly, you have my attention. I'm taking a look at SpiceDB.