All the binding names should be distinct, no repeats (since they'll all be hanging off the same env object)

[caass]

We should verify that users don't use the same binding names for

• Durable Objects
• Service Bindings
• KV Namespaces
• R2 Buckets

[caass]

Is there a list of the properties whose bindings should be checked for uniqueness? Off the top of my head, like

• Durable Objects
• Service Bindings
• Environment variables
• KV namespaces

But I feel like there's more that i'm missing

[threepointone]

R2
Also secrets, but there's no good way to validate against that without a network call

[petebacondarwin]

Something that came up yesterday in our attempt to upgrade one of the Pages Workers to W2 was the approach of setting a var binding in wrangler.toml but then overriding it with a secret in the dashboard. I don't know how prevalent this practice is, but I think we should not attempt to ensure uniqueness between vars and secrets.

[caass]

That makes sense. What about unsafe bindings as well?

[caass]

And wasm_modules and text_blobs?

[caass]

Something that came up yesterday in our attempt to upgrade one of the Pages Workers to W2 was the approach of setting a var binding in wrangler.toml but then overriding it with a secret in the dashboard. I don't know how prevalent this practice is, but I think we should not attempt to ensure uniqueness between vars and secrets.

We could do a warning, maybe, instead of a full error