Multiple Storage Buckets : Durability in buckets and specific APIs.

[asutherland]

There was a lot of discussion at TPAC 2019 about Multiple Storage Buckets. whatwg/storage#2 serves as a spec root for this, and at TPAC 2019 a small group came up with a broad strokes presentation on the subject matter.

A motivating aspect of the discussion was Chrome/Blink's intent to prototype (formerly implement) a relaxed durability transaction. The consensus of those attending the WebApps WG who expressed interest in storage and IndexedDB seemed to be that it was reasonable for buckets to support a durability/performance knob and that it was also reasonable to allow this to also be expressed at a transaction level with the same semantics.

@annevk and myself were the Mozillians most directly involved in the process.

Chrome/Blink has indicated a desire to move forward with shipping w3c/IndexedDB#50 and would like to know Mozilla's public stance on this. Since this would constitute the first step in exposing such semantics as part of the standards process, I figured it was worth filing an explicit standards-position bug on this.

Spec-wise, I guess I'd view this as marking Multiple Storage Buckets as worth prototyping and treating IDB transaction durability as related to that.

[annevk]

High-level summary of the idea: The web platform has many storage APIs to address various different/overlapping use cases. The idea behind buckets is that rather than having on large storage bucket (i.e., area) for an entire origin this is divided up. When you use a storage API you'd indicate the bucket (or you'd go from the bucket to the respective API, undecided). That helps with storage management of multiple applications on a single origin, but also allows for adding additional metadata that the browser can use in conjunction with "frecency" to free up storage space.

[annevk]

Given that this is mostly an idea at this point without a concrete form I think we should close this issue until it's further along. (Generally there's at least an explainer before we establish a position on something.)

And as far as the IDB issue goes I think it's fine to say Mozilla supports that change?

Does that seem okay @asutherland?

[asutherland]

Yes, that makes sense. I'm currently planning to work on an explainer for buckets in 2020H1 and I'll file a new issue for that when the time is right.

[annevk]

Seems this is #475 now.