Depends what you are using of it - and whether it needs to be edge based?

Like everything in a CF worker can be done on a VM or Container, it's just like you don't get the special sauce as easy, the API's pre-built to solve common edge issues.

why would you ever want to leave? welcome. to. our. cult. 😂

I agree that it depends on what you are doing, but it's not entirely trivial to self-host the code. There is no runtime available that you can run yourself, and your code won't run in Node out of the box. Maybe in the future Deno will be similar enough, but in my attempts that's been even more painful. Most of the APIs can be polyfilled, but some polyfills are quite lacking (e.g.

FormData

), and you will still need to have an adapter from a Node request to a service worker-like Request. I ended up creating a wrapper that wraps my application in Express so I could run it locally (

wrangler dev

goes a long way, but with a poor internet connection it can be painful).

Just scrolled past the above conversation, relying on something like Discord would be a very bad idea, mostly because Discord uses Cloudflare, which means if Cloudflare goes down they would also lose their main place to communicate

when communication would be of most importance

Another reason I would like to be able to run it myself without too many hacks is for on-premise deployments. My product FriendlyCaptcha has very privacy-aware customers and some large enterprises don't want to be attached to any cloud/DNS provider, so it is currently costing potential customers.

I mean, technically, I don't know if it's possible for Cloudflare to go fully offline, with the amount of georedundant servers/pops

they have, in the past?

I am pretty sure those had to do with BGP issues of other providers.

no not always, sometimes very much of their own fault as they've admitted too

And still, does not matter if its another providers fault? Just because someone else messed up on that level, does not eliminate their need to communicate so they could find the root of the issue, does it?

I completely agree. Does anyone know what slack runs on? I assume aws or gcp.

They've used aws in the past, as of right now, not certain

Was a while since I heard about it

slack.com returns an AWS IP

probably cloudfront

You would definitely have to rewrite the code, that is clear. But it's not like using CF Workers ties you in for life... and you are never allowed to leave.

That being said, they make your life easier so why would you want to leave?

Like while Fastly give you VCL which is super powerful, Cloudflare is giving Javascript API's and an Edge Computing Platform which~~ over time ~~~~should exceed~~ exceeds what VCL and the Fastly VCL extensions can do. Also, not many people know VCL..

Although, I would still love to be able to write some VCL routines in CF (but do they even use it?) - that ship has sailed 🚢

I know Fastly also have an edge computing platform - but its behind the Cloudflare one.

Don't get me wrong: I love CF Workers - they are often the perfect platform. Moving away from that platform means you have to rewrite some code. Maybe in the future that will be different, but that's how it is now 🙂

It has come up in our internal meetings when we discuss CF Workers too... vendor lock-in. I'm not too concerned because all my proposed changes are enhancements that fallback if we can't use CF Workers.

anyone configure a link shortener on workers?

yep!

want to share a guide? 👀

was just grabbing the code :p

it doesn't use an admin page like some do, it just has a simple POST to add new pages but it works for our needs