rocicorp/reflect-draw

reflect-draw's Issues

We should autodeploy the worker to Cloudflare!

It's interesting to think through exactly what should happen here.

As far as I can tell the Workers platform lacks anything like Vercel's "one-env-per-PR" feature so we can't autodeploy for PRs. I suppose this implies that vercel previews for PR only work for UI changes. Which ... ok that's a start.

I did find this Github action: https://github.com/marketplace/actions/deploy-to-cloudflare-workers-with-wrangler. We could use this to promote to production automatically, but not 100% that's a good idea?

Need to make reps resuable

Right now reps is baked into this repository. However it is entirely boilerplate except for one line - where mutators.ts is linked into the server.

We need to figure out an easy way for our customers to build apps and reuse the server, supplying their own mutators. This is complicated by the facts that:

1. CF Workers don't allow eval or dynamic import for security reasons
2. The Worker Runtime instantiates the Durable Object, not the user's app, so parameterizing the class / dependency injection doesn't work

I think that ideally what I'd like is something like:

1. npm install --save-dev reps in your application
2. the user does something (env var, runs a script ... something else?) that configures reps to get the mutators from the parent application.
3. Thereafter, whenever the user builds reps (either via npm run dev, npm run build, wrangler publish, etc) it uses the configured mutators from step 2.

Simulating client offline is tricky during development

• Using Chrome Devtools' "offline" does not work, because it doesn't work on WebSocket and now everything runs through WebSocket
• Accessing localhost from another device one the network does not work, because SubtleCrypto thinks it's insecure and will return undefined for all its fields.

For now the only way I can think of is to turn off and then turn on the Miniflare dev server. But it's not the same situation as some client going offline.

Is there a better way?

Question on init mutations

https://github.com/rocicorp/replidraw-do/blob/d4f191ea47a885636214db1451feccac807d25d2/src/pages/d/%5Bid%5D.tsx#L47-L56

Why is it OK here to call initClientState directly, while initShapes is delayed until after the first sync?

(Also, LMK if this is/isn't a good place to ask questions)

Rip out current scheduling code - it's not working well

We are experiencing a problem with this server where the peer-to-peer latency is unexpectedly high: you move a mouse on one tab and it takes awhile to show up on other other tab. Watching the server output I can see that what's happening is that something is going wrong with scheduling: the server calculates that the frame should play in the future and waits -- sometimes awhile! -- to play it.

It seems to get worse the longer the client is connected. I regularly see the server wait a full second which makes the experience pretty bad.

Cloudflare Workers have hobbled clocks for security reasons: https://developers.cloudflare.com/workers/learning/security-model#there-is-no-fix-for-spectre. It seems very likely that the behavior observed is related to this somehow, but I can't figure out what's going wrong yet.

Not working on Safari (both mobile and desktop)

It only works when using Miniflare. But once deployed it does not work (try https://replidraw-do.vercel.app in Safari) and the console says:

[Debug] roomID=LWyzh- req=p4twbolFsCwS-qfRaB9Y5 – "Skipping duplicate connect request" (598-0f8f78f483115daa.js, line 1)

Could be something on how Safari deals with WebSocket and https?