ie, it would be cheaper than paying for wall time

Given that it's free to proxy websockets, I'm hoping that (at least) the paid workers plan will bundle a generous amount of websocket termination. My pet project is a (self-funded) distance-learning resource for teachers and some of the proposed billing setups will rule out Cloudflare for me!

I am hopeful as well.

I think in the end a reasonable pricing model will prevail 🙂

Fingers crossed! I doubt that the orange names we see in here make the decision so hopefully those in control of the purse strings agree 🙂

Any ideas about how to use durable objects for managing real-time state across multiple different objects of interest?

Ie, a client might want real-time websocket updates regarding its user data, game data and chat data, which are all separate durable objects

But it seems a little awkward to have the client open a ws connection for each

is it possible to access a Postgres instance from a Durable Object? Would the DB connection hang around?

Hmm, I have a single instance of a DO that hangs indefinitely on any call to

this.state.storage

(list, get) no matter how trivial. The DO instance is alive, it responds to other non-storage requests. Any way to further debug or access the state in another way?

Try a “Comm DO” object and send the updates from other DO to this one. And the user will only need to connect to that end point.

How is that call? Using await, promise.resolve?

await. The promises never return, and the http call never times out! It feels like a db transaction lock honestly, but I'm not using transactions.

Probably a unhandled exception, but only cloudflare devs can look at that!

hmm, I see what you mean

a muxer sorta?

is there a way to specifically allocate a DO to a region or no

Maybe. A simple DO that you can use to concentrate and dispatch all messages are e enough. If you users are DO too maybe just dispatch all messages to that DO can do the job.

Only EU for now.

gotcha

and uhh, will multiple WS clients be connected to the same worker for a DO

or not necessarily

I was hoping to have all users of a region connect to a worker in that region who could proxy/muxe all the data

The DO will be first created close to the first used that trigger the creation. All other at same region probably will be connect on the same. You can look at cf headers and create different DOs based on the region by yourself.

Alright because ideally I can have just a few durable objects per region

Because I’m hoping to basically do muxing to reduce bandwidth in my network

And was hoping to use the edge network to perform the muxing

But muxing is only effective if you can get multiple users in a region to connect to the same worker/DO

Another think that i'm missing is the ability to the DO know his own ID (preferably the used on "fromName"). That's useful when we need to fetch external resources as part of the initialization, from KV or other source.

Hi. Is there any example of creating a auth/rbac service in Durable Objects? I would like to store the user's token and their permissions in this. I can imagine how to do this. But an example could help a lot thx