although at this point, the caches should have expired

all read calls in SpiceDB (check, expand, read, lookup) by default use "best effort" consistency to ensure performance, which means they can be partially or completely cached for a time

Thanks. I will read up on that. Does each "thread" have its own cache?

no, but each instance does and some may not hit the cache before others write to it

we're going to be adding these consistency options to

zed

soon, but in the interim, you can add them to any API call you make via a client library

ok, thanks!

we generally recommend using

fully_consistent

for local testing, because its the most "expected" output; for a real application, you should only use it if you absolutely need up-to-the-second consistency; otherwise, the recommendation is to send in zedtoken to safely use caching where applicable

yes, I just read through the doc you sent over and that makes sense.

ye aware of that, but thats not an issue for our use case, that we read stale info a short period of time (a few seconds)

@User just be aware it is an unsupported setup, so it may break in unexpected ways as well

for example, if you send a zedtoken in and the read replica doesn't have the revision available, the API call will fail with an error

I mentioned this on https://github.com/authzed/spicedb/issues/161, but has anyone tried out the Postgres interface for Spanner and if SpiceDB would work with it?

@User we have not, as Spanner just added support for it

yes, it's very new

if Spanner does support "full" Postgres dialect, it would likely work

we do test that driver on CRDB too

flattened graph watch api design

do we have example of how google use zanzibar in GCP? we have similar use case in our company internal authz.

@User you mean to manage cloud permissions?

yes

we will be writing a blog post about it soon

okay

do you mean practically applying spicedb to gcp in some way or writing a schema similar to the one google uses to power Google Groups?

Can we run multiple instance of SpiceDB and with external DB will it keep the instances in sync ? Just want to know if spiceDB is stateless?

I just want example schema, like we have google doc schema in playground

We are evaluating to move from our handrolled Zanzibar impl using Oso to AuthZed. But the whole project is on AWS Lambda using TypeScript. I don't really care about v0,v1 too much right now because I'm stuck fighting TypeScript to spit out declarations for the project. If you have any at hand, it would be very helpful.