Blocking: split deployments

[boehs]

Pinefore was initially deployed from one codebase, where vite was responsible for managing everything. This configuration proved problematic for the following reasons:

• f/e shims changed b/e behaviour
• vite sometimes would do security things backend doesn't need
• very likely, there is redundant code via the bundle, potentially also impacting startup times
• package.json was getting unwieldy

This all came from frustration with how vite handles wasm, but that was just what pushed me over the edge. the following is gained by deploying separately:

1. very likely less errors
2. potentially faster by a couple ms
3. separation of concerns, which reduces the bug surface for attack

Cons:

1. Huge rewrite introduces bugs
2. Separate deployments means one deployment could fail without the other
3. Workers is a much more manual deployment process, which in some ways is nicer

Todo:

• [x] Separate both projects into their own workspaces
• [x] Copy tokens
• [x] Get API deploying
• [x] Get F/E deploying
• [x] Connect in SSR
  • [x] ish
  • [x] make query client use binding instead of fetch
• [x] Bugs
  • [x] Some API routes do not work
  • [x] Logged out?
• [ ] DX
  • [ ] Figure out localhost situation for F/E

[linear[bot]]

PFW-42 Blocking: split deployments

[boehs]

Casualties tonight:

• 10 hours
• 20,000 lines of code