4 ms·
I don't care about how big a page is. I care how fast it loads. And some of these pages fail that test badly. For example, sdf.org takes a whopping 1.60 se
by londons_explore 4y ago
I don't care about how big a page is. I care how fast it loads.
And some of these pages fail that test badly. For example, sdf.org takes a whopping 1.60 seconds for the home page GET request to return a single byte.
I'd like to see a new leaderboard of 'instant' pages. An instant page is one where a typical user click has the result fully rendered inside 100 milliseconds (almost imperceptible latency).
- speed_spread 4y agoI'd rather have a page load in 2 seconds and be done with than one that shows up in 500 millis and keeps loading advertising shit in the background for 10 seconds. Page size is merely a hint about the level of counter-user sophistication, the lower the better.
- yoyohello13 4y agoI’d rather a page that loads in 500ms and be done than either of those.
- onion2k 4y agoAn instant page is one where a typical user click has the result fully rendered inside 100 milliseconds (almost imperceptible latency). That's an incredibly hard to reach target. Establishing a secure connection from the browser to a server requires several round trips - 1 for DNS lookup, 1 for TCP handshake, 2 to establish the TLS connection, and 1 for the HTTP request. If you divide your 100ms perf budget equally between them that means your servers can be no more than 1800 miles away from the user using (100ms c) / 2 (for the round trip) / 5 requests*. Using TLS 1.3 removes one round trip which is definitely a good optimization. That's before you've even started any serverside rendering, session handling, db connections, etc. Even with a really good edge network that's going to be incredibly hard to achieve. I'd love it if every website served a page in 100ms, but it's never going to happen. I got https://ooer.com https://ooer.com down to below 500ms for myself (495ms last time I checked) after a fair bit of effect. Getting any lower would be a waste of time. Users can wait an extra 400ms.
- ryanisnan 4y agoI recommend the parent reads High Performance Browser Networking for a good overview of the latencies involved. Thanks for breaking it down here!
- londons_explore 4y agoThe 'typical user click' doesn't have to be the first page load... Depending on the application, many clicks might be from page to page in the web-app. And it's very possible for the majority of clicks to involve no server roundtrip at all.
- onion2k 4y agoHN readers will get quite cross with you if you recommend clientside rendering.
- initramfs 4y agoMy page loads in 174ms using GTmetrix.com, but I also don't really have a lot on my page. It's more of an experimental test. I was going to submit mine to their club but i don't have 100 points for mine yet https://news.ycombinator.com/item?id=33686805 https://news.ycombinator.com/item?id=33686805
- kingofpandora 4y agoMaybe I can't afford a powerful server. But I can always afford not to use 6mb of JavaScript.
- serf 4y ago> I don't care about how big a page is. I care how fast it loads. you should care about both, not everyone has unlimited data capacity.
- neurostimulant 4y agoThis seems to be really hard to do unless you're using global cdn and forgoing https. Ping between US and SG is almost 300ms in the best condition. That alone already blew the 100ms budget for visitors far away from your server location. And then we have https, where initial https handshake alone could take several hundreds ms (depending on various reasons).