I think the thing that confuses people the MOST about Core Web Vitals is:
When their lab data tells them things are good, but their field data says they have a problem… 🤔
So I wrote a post to help explain why this can happen and what to do about it: https://t.co/YqcvsiCgrE
@NicolasComet @Fabinout L'un de ses 3 la, et bien d'autre. C'est en pensant qu'il y a une voie "naturelle", à la pokemon, que l'on construit des carrières malheureuses. L'experiences s'expriment de bien des manières.
protip for JS devs: never trust transpilers, always write tests that run AFTER building/compiling ...
testing your source code is not effective, since that's not the code your users are running.
@jrychter As individuals this is sad indeed. As a society I guess we do since... it is the case (being cynic here). Reciprocally, you could ask also do we want to have very rich people around? Both come together.
@gcouprie@clever_cloud 3/ Pair debate here, I'm trying to understand what am I missing from the picture. Thanks for helping me on this. Also cuddos for using protobuf, it surely reduces the footprint. I would love to see some benchmarks.
@gcouprie@clever_cloud 1/ ty :) No hard feelings here. Reading the git live, impressive work, cool to have it oss'ed. I can elaborate (from my cur groking):
Biscuit moves the ACL logic from the auth layer to the token, on cost of heavier tokens (?).
@gcouprie@clever_cloud 2/ At opposite, keeping the acl logic in each client (you mention uservices in the blog, I assume you do have chassis in each of them, so easy to ship this logic) and lightweight tokens with just the claims would achieve the same (acl are just kv and some ifs).