Russ Cox

> Actually, looking again this isn't the same thing. The next call to Token would return the start token, but in the initial examples the start token has already been...

d.Token would return one, but that's not how you'd use d.Push. Instead you would make a new decoder, as in your original comment above, and seed it with "one".

It is a design goal for OSV for all clients to be able to determine in an algorithmically precise way whether a particular version is affected. Having "database_specific" in the...

I agree with @oliverchang about being very selective in how many different constraints we add, for all the reasons he enumerated. I suggest we start with just "last_affected".

I believe it is too late to make a backwards-incompatible change like this. It would be fine to say in prose that duplicate entries for a given type should not...

> Regarding format/version declarations I think we should support this because we support namespaces What does it mean that we support namespaces? That's surprising to me.

For clarity, here is the implementation of Do from inside Google, which makes the semantics a bit clearer: ``` // Do runs the function f as allowed by First, Every,...

Based on the discussion above, this proposal seems like a **[likely accept](https://go.dev/s/proposal-status#likely-accept)**. — rsc for the proposal review group

No change in consensus, so **[accepted](https://go.dev/s/proposal-status#accepted)**. 🎉 This issue now tracks the work of implementing the proposal. — rsc for the proposal review group

Let's not drag finalizers in. If this turns out to be a real problem, we should probably think about undoing the change. @aclements points out that we could also do...