Can the window parameter be 0?

[Acconut]

According to https://www.ietf.org/archive/id/draft-ietf-httpapi-ratelimit-headers-08.html#name-window-parameter, the window parameter is a non-negative integer, meaning that it can also be 0. Does it make sense to have a window size of 0 seconds? Should this be allowed or should we require it to be greater than 0?

[ioggstream]

iirc the rationale was that w should be compatible with delay-seconds https://datatracker.ietf.org/doc/html/rfc9110#section-10.2.3-4

This could be reconsidered, though I remember we (long time ago) used non-negative intentionally :)

[Acconut]

iirc the rationale was that w should be compatible with delay-seconds https://datatracker.ietf.org/doc/html/rfc9110#section-10.2.3-4

The Retry-After fields appears more similar to the t from RateLimit. For that it makes sense to have 0 as a value, so a client can retry as soon as possible. My comment above was referring to the w from RateLimit-Policy, which identifies the time window of the policy. Are there use cases for empty time windows?

[darrelmiller]

@Acconut I think there are two ways to interpret w=0. It either means that the quota is replenished after zero time, and therefore there is effectively no quota. Or it means there is no reset window and so w=0 is equivalent to not providing a w parameter at all. This can be used for the "trial" scenario.

I suspect rather than have this ambiguity, we should just make the value a positive integer greater than zero.

[Acconut]

I suspect rather than have this ambiguity, we should just make the value a positive integer greater than zero.

I agree. That makes more sense for such a document.