Life-hack: Tilføj '*.queue-it.net' til din adblocker.

RT @andreas_madsen: After my blog post on getting a Spotlight award at ICLR, as an Independent Researcher, I got nearly 2000 emails. Many h….

RT @finnpauls: @emilbayes

Try running any kind of non-trivial code on the EVM at these fees! Combination of high Eth exchange rate and extremely busy network, driving up fees. $0.60 before you've even executed your first instruction

Nevermind, I was conflating "ReplacerFunction" and "PropertyList" in the spec.

Update my old module to take a replacer function, and think I found a bug in Chrome and Firefox? replacer should work on arrays, when you pass replacer as an array, both skip this step? eg: JSON.stringify([0, 1, 2], [0]) // should be [0, null, null].

Another neat bit hacking trick; rounding a number up to a next power of 2. Useful for aligning pointers (eg when slicing a Uint8Array as a Float64Array). Also works in plain Javascript. Code can be found here:

Code can be found here:

Fun bit hacking on implementing the missing bswap instruction in WASM (eg. converting LE to BE and back). Is there a more optimal way? I guess one get and set could be replaced with a tee

Wasm gives us very predictable performance and i64, even if we pay a penalty by having to memcpy between each invocation. The above perf numbers are including memory.set!.

We specifically hit this limit in sodium-javascript, which is synchronous, sinc the async scheduling simply kills perf. For hashing we’re anywhere from 80% - 400% faster than native APIs, since we have hand written our wasm. Our modules are already highly optimised in byte count.

Streaming, async parsing and compilation is a great feature, but if you want to use wasm anywhere in your dependency tree, the async nature has to leak all the way to the top. Mind you this limit is Chrome specific. It’s not spec’ed, and all others have no or much higher limits.

Nothing stops you from loading 10’s of MBS of ES2020, with a wildly contextual grammar and extremely complicated parsing, causing your browser to stall for 100s of ms, while you can just load 4kb of unambiguous, straight forward WASM, without having to resort to async loading.

RT @andrewosh: Very excited to announce Hyperdrive v10! . Better indexing, snappier networking, a daemon, FUSE, and a bunch of other cool/n….

RT @mafintosh: Distributed holepunching through the Hyperswarm DHT

Also, there's no point in using ES384 or ES512 unless you need to be compatible with another system. Those algorithms use very different cryptography that is not worth it. Stick to ES256.

If you use ES256 instead of RS256 on your JWTs, you save 75% on signature size.

RT @paoloardoino: If you're a great #NodeJS dev and want to code for @bitfinex this is your chance!.

RT @hyperdivisiondk: After a lot of hard work from from @chm_diederichs, we released a new major version of sodium-native@3.0.0 today with….