We've seen some internal analytics initiatives take place that tend to use lookup/lookupsubject semantics

(we call this internally "actor enumeration" and "resource enumeration")

Thanks a ton for the feedback @User ! 🙇🏻‍♂️

you're welcome! i'll be back in about 90 minutes

Good morning y'all 👋🏻 I know authzed folks built the quay.io, but: are there plans to push spicedb's container image to GitHub's container registry? Y'all are pushing releases (https://github.com/authzed/spicedb/releases) there, so I'd make sense to also have the image available. WDYT?

Hi friends, have you tried spicedb for a huge amount of documents? say 30 millions in total, and the documents are organized in hierarchy(like a directory tree), with inherit permissions. The challenge is to build the tree according to the permissions for the end users. I have tried in development model with around 400, 000 documents. I was using lookup trying to retrieve the permitted documents, it takes around 2 hours😅 . and for sure I must be using it in the wrong way, but do you have any suggestion on this topic?

We will talk about it today and get back to you. We don’t do it for our production images because the pricing is oppressive, but for release images it might be fine.

We haven’t gone that far yet. How deeply nested was your test with 400k? 2 hours is obviously not acceptable.

Also how connected is your graph?

Hi Jake, thanks for the reply! The testing data was a binary tree, I set a write permission for a user(say shiqinwen) on the root node, and other node inherit the write permission, then I tried to lookup all permitted nodes.

This is actually a real case in my product with is currently a knowledge base around 100K people using it daily. The permission is very flexible(due to the inherit permission), and we spent so much effort on optimizing the loading speed. Until I saw Zanzibar and realized that we was not even speaking the permission domain language in our system. I like spicedb a lot, it's simple and very expressive. So I'm thinking if I can adopt it in my product😀

Can you share the test with us?

sure, just a sec

The code is written in Java, runs on spicedb development environment. and sorry the code looks cumbersome....😅

No problem, we will take a look today and get back to you with findings

What do you mean when you say development environment? Do you mean the developer api?

ah, I ran spicedb locally on my Mac laptop

🤔 According to the billing documentation, it's free for opensource? >GitHub Packages usage is free for public packages. https://docs.github.com/en/billing/managing-billing-for-github-packages/about-billing-for-github-packages

Sounds good! We briefly considered it for our production Authzed.com deployments which have many private and vendored images, that’s what was too expensive.

oh yeah, for private packages is certainly not free. 😅

@User did you run with spicedb at head or did you use a released version?

I seem to be locked out of my GitHub tenant

it a release version

There are some improvements (read: caching) to lookup that haven't made their way into a release yet, try that and see whether it improves your lookup performance

@User looking...