Hacker News Reader

Comment by flukus | Hacker News Reader

> Now, if you just dumped your other ad networks and ran everything through us, I bet it would be load much faster and that badge might magically disappear...

My experience with everything google has touched lately suggest that this wouldn't improve speed any. Gmail and youtube make continental drift look speedy and even the search page takes 1.4MB and takes over a second to load for me (maybe corporate network issue), that's approaching the size of doom to display a dozen links.

Google doesn't have any moral authority when it comes to bloat.

Edit - for reference HN takes about the same time to load, but that has to cross the pacific ocean whereas google supposedly has local data centers.

View on HN

I think the point being made is that it's no longer actually about the speed, it's about the 'badge of shame'. The analogy being used is accurate, in that 'protection money' isn't, and was never, about protection.

Part of the problem is that they place high importance on the first contentful paint (aka FCP) above all else. They're not very much looking at how it performs overall, mostly just how long it takes for a user to see the first bit of content. If that were a bit of text that said "loading" with a spinner that goes for a minute or two, they may consider that fast.

This section shows how Lighthouse calculates your overall Performance score: https://web.dev/performance-scoring/#lighthouse-6

(Note that the table gets cut off on narrow mobile screens... I’ll file a bug)

FCP is 20% of the score. The other metrics capture different milestones in the loading experience. For example Total Blocking Time is intended to bring awareness to sites that may look complete but can’t respond to user input (because the main thread is busy running JS).

The metrics overview that we just launched provides more detail about how our metrics were designed to capture the end-to-end loading experience (or at least that’s what we’re working towards): https://web.dev/user-centric-performance-metrics/

So in the case of your site-that-just-loads-a-spinner example, yes it might have a good FCP time, but it’s LCP time probably wouldn’t be that good, and therefore the overall Performance score would be mediocre.

Long story short, I don’t think it’s as easy to game a good Performance score as you might think

Disclosure: I work on web.dev

The only PC where I saw Gmail being fast was a Ryzen 9 3900x build, I suspect that if Google devs were given shittier PCs they'd build faster products that would appeal more to the average user.

I run a Core i5-2500K from 2011 and Gmail runs just fine. Far better than Thunderbird or Outlook ever did, but it does take a lot of memory.

I have a Google-issued corporate workstation with 64 GB of RAM and 12 cores. Gmail is really really slow. Like so slow that I if I have to refresh it I'll go get coffee. But I'm on Firefox, and so "I Am Not The User" and all that.

IMO the best way to use Gmail is to open and pin it, only load once a day. It seems optimised for this use case.

There's a big difference between "done loading" and "appears to be done loading". Google is reportedly very cautious about the latter, and pulls some hijinks to appear faster than it is. Frankly, I'd be happier if more sites paid a similar level of attention.

That said, if your search page takes over a second, you're right, it might be an issue with your network.

> that has to cross the pacific ocean whereas google supposedly has local data centers

HN is behind Cloudflare, and so also has local DCs

Reply on news.ycombinator.com