dev meet

meeting-notes/2024-04-14-dev-meet.md

@@ -0,0 +1,58 @@
+# 2025-04-14 Sync Kiara/Valentin
+
+- @fricklerhandwerk asked for a walk through https://git.fediversity.eu/Fediversity/Fediversity/pulls/307
+    - @kiara:
+        - TF seemed like easier to work with
+        - A lot of details, but got it to run with the deployed infra
+        - Tried TF with nixos-anywhere, but that [seemed not to pick up on our config](https://git.fediversity.eu/kiara/Fediversity/pulls/1)
+        - Also tried [terraform-nixos](https://github.com/nix-community/terraform-nixos/), but that is deserted
+- @fricklerhandwerk: probably need to take a few steps back since the project is doing a backflip, and re-assess what we want morally and then re-align on the technical strategy
+- @kiara desired work order
+    - website
+        - remove the unused code
+    - panel
+        - should be a reference implementation to demo the ideas
+        - let front-end people productise independently, using our APIs
+    - infra
+        - want to replace NixOps4 with Terraform
+        - we have private keys lying around in the repo
+            - @fricklerhandwerk: can just move them to agenix?
+            - @kiara: yes, need to nuke everything and start over
+    - pins
+        - need to consolidate the different ways of doing things
+    - VMs
+        - eventually want to remove the hard-coded stuff
+- @fricklerhandwerk: what about integration tests? our basics is not tested, we need the proxmox workflow under control before doing other stuff
+    - @kiara: agreed. would like to get to a system that others can pick up by reading through the tests
+- @fricklerhandwerk: possibly we need to take even more steps back, such as nailing down our development workflows
+    - e.g. we can't even make suggestions in code reviews on Forgejo
+    - @kiara: we could spin up Gerrit, but where does it stop?
+        - @fricklerhandwerk: outline:
+            - open up an infra repo (or eat the existing one from within)
+                - set up SSO, some Git server, issue tracker, and a deployment workflow for that
+            - start over with the application code, begin with Nix first, integration tests first, no PR accepted without tests and reviews
+                - spiral up from there; we already know a lot of the rest of the owl
+            - it may end up amounting to bootstrapping the whole infra idea...
+        - @kiara: agreed, the basics aren't testable because everything is hard coded
+    - most importantly, everyone on the team should be able (i.e. capable and allowed to) change anything about the system, and for that we really need to agree on how it works and how to approach things
+        - which is also why we'd need a proper architecture decision record (ADR) system, by which I simply mean being disciplined about writing things down
+- @kiara: I consider the application services and the deployment code around it the core of our work
+    - the (currently Django) shell invoking that I consider an example implementation of the product on top
+        - we already know there will be other such implementations because Procolix wants to productise it in their existing setup
+        - getting the boundaries between those clear before new people (e.g. external contributors, future team members) come in
+    - therefore the most important thing would be building an interface for spinning up the Nix environment and invoking the deployment from wherever
+        - currently it's only documented as embodied in our implementation
+    - @fricklerhandwerk: agreed; and we should reduce our scope for what we work on to the absolute minimum, and focus on user stories for integrators
+        - although they should still be expressed through something you can click on, based on automatic testing of course
+- @fricklerhandwerk: it seems we agree that we need to do a spring cleaninig
+    - @kiara: yes (service modules seem not to be the problem though)
+- @fricklerhandwerk: would terraform not make things harder?
+  - @fricklerhandwerk: agree it's not as mature still tho
+- (some back and forth on Terraform vs NixOps4)
+    - @fricklerhandwerk: although the decision to stick with NixOps4 was made a long time ago, the rationale seems not to have been written down
+    - trade-offs:
+        - NixOps4 is not mature and can't do some crucial things yet; may risk the timeline
+        - Terraform done from json/nix could use Nix language wrappers
+            - TF references are stringly typed, it's slightly brittle
+- next steps:
+    - start pairing tomorrow on the spring clean, and shovel code