4 ms·

I agree completely regarding "users want fast enough". Also: for regular sites, html/css/js optimization is less important than server location. If your server

I agree completely regarding "users want fast enough". Also: for regular sites, html/css/js optimization is less important than server location. If your server is in Europe and your user is in the US, that's the big one, not your HTML, CSS or JS. And if you're fetishizing over Lighthouse scores, stop. It's only a very rough measure and shouldn't be treated as a goal itself.

pastage 4y ago
I agree if the website does lots of new connections with no cache, the users have a slow last mile internet, and we are talking about LA to Moscow. Otherwise trans Atlantic is fine, 80-150ms max ping is not that bad. EDIT max ping of each of our recurring American users on websockets, about a thousand of them. No difference in usage pattern there between Europe and US, but sure we do not have many that have worse ping.
- luckylion 4y ago
  For us, effective page load time increases annoyingly when you're more than 1500km from the server. Fortunately most of our traffic can easily be cached, so having a caching server on US East and US West works fine.
lelandfe 4y ago
> It's only a very rough measure and shouldn't be treated as a goal itself Lighthouse is a synthetic testing tool. It can be incredibly useful when you connect it to your CI build process and track scores over time – revealing how code changes affect (some) front-end performance metrics.[0] Of course, some scores are non-deterministic so you'll need to run it multiple times.[1] And you should be tracking real user data as well to validate it. Using PageSpeed Insights to benchmark a site is akin to loading your website a single time on a single device and saying "wow, my site is slow" Using Lighthouse in your browser to benchmark a site is akin to wasting your time. [0] https://github.com/GoogleChrome/lighthouse-ci https://github.com/GoogleChrome/lighthouse-ci [1] https://developers.google.com/web/tools/lighthouse/variability https://developers.google.com/web/tools/lighthouse/variabili...
withinboredom 4y ago
If your entire page can fit in a single packet, RTT barely matters for most of the earth.

pastage 4y ago

I agree if the website does lots of new connections with no cache, the users have a slow last mile internet, and we are talking about LA to Moscow. Otherwise trans Atlantic is fine, 80-150ms max ping is not that bad. EDIT max ping of each of our recurring American users on websockets, about a thousand of them. No difference in usage pattern there between Europe and US, but sure we do not have many that have worse ping.

luckylion 4y ago
For us, effective page load time increases annoyingly when you're more than 1500km from the server. Fortunately most of our traffic can easily be cached, so having a caching server on US East and US West works fine.

lelandfe 4y ago

> It's only a very rough measure and shouldn't be treated as a goal itself Lighthouse is a synthetic testing tool. It can be incredibly useful when you connect it to your CI build process and track scores over time – revealing how code changes affect (some) front-end performance metrics.[0] Of course, some scores are non-deterministic so you'll need to run it multiple times.[1] And you should be tracking real user data as well to validate it. Using PageSpeed Insights to benchmark a site is akin to loading your website a single time on a single device and saying "wow, my site is slow" Using Lighthouse in your browser to benchmark a site is akin to wasting your time. [0] https://github.com/GoogleChrome/lighthouse-ci https://github.com/GoogleChrome/lighthouse-ci [1] https://developers.google.com/web/tools/lighthouse/variability https://developers.google.com/web/tools/lighthouse/variabili...

withinboredom 4y ago

If your entire page can fit in a single packet, RTT barely matters for most of the earth.