Victor Costan

I think that what you want should be covered by the [Web Share Level 2 API](https://w3c.github.io/web-share/level-2/). I don't know if any browser implemented Web Share Level 2 on desktop. For...

FWIW, not having a close will make benchmarking more difficult. This is not a core use case, but I think it still matters to the API's long-term well-being.

Your last sentence nails my intention -- thank you for clarifying this! 1) IDB is fairly widely deployed at this point, so the changes I've outlined may not be Web-compatible....

## Memory Cache I agree with @domenic that an in-memory cache is complex enough that it deserves its own API layer. localStorage doesn't give any cross-tab synchronization guarantees. The price...

@esprehn Thank you very much for the detailed explanation here! Risking over-simplifying, I claim that you're advocating for a local optimum. I read your comment to say you'd prefer a...

Sorry, I have no idea how I missed #2. Would you prefer that I close this and comment there? I haven't opened an issue for values yet because I'm still...

Having written the above, I'm starting to think that I'd rather that we limit key sizes instead of types. What are your thoughts on the two axes?

Answering the last question: key serialization is implementation-dependent. To be consistent across browsers, we'd have to define some cost model and use that. We can have a simple and efficient...

compare-and-swap seems like a specific case of read-modify-write. If we do add a primitive, I'd prefer that it's `readModifyWrite` (or `map`?). It would start a txn, read a key, call...

`overrideMimeType` is mentioned as a hack. I would suggest implementing `responseType = "arraybuffer"` instead ([MDN reference](https://developer.mozilla.org/en-US/docs/DOM/XMLHttpRequest/Sending_and_Receiving_Binary_Data)), because [node.js supports typed arrays](https://github.com/joyent/node/issues/1322).