@rysiek@mastodon.technology This seems to actually be an HTTP/1.1 vs. HTTP/2 test, since no major browser supports HTTP/2 on unencrypted connections.

@rysiek@mastodon.technology It's basically an artificial limitation, meant to push HTTPS. But I'm fine with that...

@alexandra @rysiek I'm cool with it because as soon as Google falls, Google-built specs like HTTP/2 and /3 will probably go with it.

@ocdtrekkie@mastodon.social @rysiek@mastodon.technology Honestly I doubt HTTP/2 and HTTP/3 will go anywhere unless Web applications get a lot simpler and lighter-weight.

@alexandra @rysiek However long they last, I guarantee you HTTP/1.1 will last decades longer.

And I think lighter weight web applications is inevitable, the current garbage is obscenely poorly engineered.

@ocdtrekkie @alexandra @rysiek You'd also have to get rid of Cloudflare, Amazon, Nginx, LiteSpeed, Facebook, Mozilla, and all the other QUIC implementors. And even then, a few community-built implementations will still remain.

@Seirdy @rysiek @alexandra I mean, without Google unilaterally deciding to handle 75% of the world's web traffic with it, I suspect it will naturally fall out of favor.

But ultimately, HTTP/1.1 support can never really be dropped, but HTTP/2 and /3 is really only doing favors for select big guys, and can be trivially replaced with HTTP/4.

@ocdtrekkie @rysiek @alexandra Without Google unilaterally handling that much traffic, Netflix + Amazon + Cloudflare would. Amazon (+Twitch) and Cloudflare were early adopters that developed and pushed their own implementations and Netflix would likely make use of it as well.

Any site with a very large number of resources per DNS lookup would benefit from HTTP/2; this includes the adtech-powered Web. Google is one of *hundreds* of companies in this space; if Google disappeared, people would sign onto a different provider.

So incentives to use HTTP/2 and HTTP/3 are still there without Google.
@ocdtrekkie @alexandra @rysiek What you need to understand is that for companies like Amazon, Netflix, Zoom, hell even Pinterest (have you seem the network monitor on that website?): multiplexed low-latency connections are what they built their entire stack around. Yes, they gracefully degrade to HTTP/1.1, but that's not the protocol use when making their most important benchmarks.

Moreover, tons of web apps don't bundle resources efficiently or stream multiple requests with a concurrency model that combines best with multiplexed connections.

There's too much non-Google money in HTTP/2 and HTTP/3 for them to be going anywhere.

If you don't believe me, cross the Great Firewall of China and inspect your network monitor. If Google goes, HTTP/2 and HTTP/3 adoption will still increase. There's too much money to be saved given those benchmarks; organizations won't sign up for an HTTP/1.1-only CDN or hosting provider.
@alexandra @ocdtrekkie @rysiek One last thing: what about non-megacorps? Well, they're still beholden to the whims of other megacorps.

If Google Search goes away, Bing and Bing-based engines (Yahoo, DDG, Ask, You, etc.) will take its place. Google uses page load performance as a ranking factor. Guess who else does? Bing.

So everyone will still have a powerful incentive to squeeze out those milliseconds from their load times but also have an incentive to monetize everything they can, including user attention spans. Most browsers (not just Chrome, but Safari and Firefox too) only accept Brotli-encoded content over HTTP/2.

So HTTP/2 looks like an obvious choice to most orgs. All the incentives are in place. I don't really see how HTTP/3 will be different when all major providers, servers, and browsers support it.
@alexandra @ocdtrekkie @rysiek TLDR: from the perspective of people financially incentivized to deliver many assets (eg ads) on a single page, there is a financial incentive to use HTTP/2 with or without Google. It either comes down to saving money at scale (big distributors like Amazon or Cloudflare) or SEO (Bing also rewards load times).

Thank you for coming to my TED Talk.

@rysiek honestly, results are very flaky here, doesn’t seem to mean much at all

but any excuse for more encryption is good :blobcatuwu:

@rysiek here it’s like, HTTPS(/2) is ~0.6 and HTTP(/1.1) is ~0.8, but quite often both might hang and HTTPS might get a lot slower than HTTP as well so :blobcatgoogly:

@rysiek Hmm, encryption wouldn't speed it up. I think the speed difference is because they're serving HTTP/1.1 unencrypted and HTTP/2 encrypted.

@rysiek@mastodon.technology That's a scam. The unencrypted version uses HTTP/1.1 and the encrypted version uses the faster HTTP/2.

@tastytea I wouldn't call it a scam. It's a bit misleading, sure, but then again: show me a browser that can do HTTP/2 without HTTPS?..

So *technically* they're somewhat right: enabling HTTPS on your website will make it load faster.

@rysiek@mastodon.technology “Encrypted Websites […] are Significantly Faster” — lie. They can be faster when HTTP/2 is used, under some circumstances. For example on sites with a lot of small pictures. It doesn't matter much on an average blog or when most of the assets are cached.

You could just as well argue that bicycles are a lot faster than cars, because the cars are stuck in traffic jams on some roads.

@rysiek Now do the same comparison with a lot of caching proxy servers along the way, like the World Wide Web was originally conceived.

Sign in to participate in the conversation
Mastodon for Tech Folks

This Mastodon instance is for people interested in technology. Discussions aren't limited to technology, because tech folks shouldn't be limited to technology either!