Re:Multi-core cryptography for Node.js
by move-on-by on Thursday April 26, 2018 @08:36AM
(#16931264)
> To be clear, I'm not saying that support for async crypto isn't worthwhile, but I'll strenuously object to making it the default without extensive evidence that it doesn't introduce performance regressions.
Pretty basic requirement for making performances changes: provide the actual numbers, or GTFO. He was asked multiple times to provide some numbers, sometime indirectly, and then directly in the above quote. Each time he responded with speculation:
> My bet would be the async crypto that's running on 8 cores.
> what will the latency look like [...]? It's actually going to be through the roof. The op will have been waiting around for a few seconds and the op's latency will be incredible.
Missed opportunity to back yourself up with facts, even if the situation is contrived to be the best-case scenario.
Pretty basic requirement for making performances changes: provide the actual numbers, or GTFO. He was asked multiple times to provide some numbers, sometime indirectly, and then directly in the above quote. Each time he responded with speculation:
> My bet would be the async crypto that's running on 8 cores.
> what will the latency look like [...]? It's actually going to be through the roof. The op will have been waiting around for a few seconds and the op's latency will be incredible.
Missed opportunity to back yourself up with facts, even if the situation is contrived to be the best-case scenario.