8 replies

1. Zach Leatherman 30 min later

   Intuitively, these charts might tempt you to compare these hosts, making assumptions about the investment each of these services put into their server infrastructure — but that wouldn’t be the whole story.

   Knowing that TTFB is both server AND client-dependent (including times for Service Worker init, DNS, TLS), this graph might instead represent how wealthy each service’s audience is.

   1. Zach Leatherman 33 min later

      We need better web performance metrics.

      1. Khalid ⚡ 37 min later

         @zachleat I thought I was looking at the new Joy Division album cover 😅

      2. Zach Leatherman 40 min later

         For me this also highlights the incredible sleight of hand that can occur when a metric swaps from lab to field data.

      3. Andy Luhrs 48 min later

         @zachleat Like what?

         1. Zach Leatherman 1 hr later

            @aluhrs maybe a server response time metric that isn’t dependent on the client so we *could* draw at least some conclusions about hosting performance or backend performance

      4. iamdtms 58 min later

         @zachleat Okay, but it's hard to reach 100% already.

         1. Zach Leatherman 1 hr later

            @iamdtms variability is part of the reason!