The book says 7% of all internet traffic already uses QUIC (HTTP/3) and Chrome has long implemented Google’s version of it.
But this book isn’t about concerning yourself with using it or implementing it, it’s about understanding what the future holds, how it works, and what roadblocks lie ahead.
The lack of API support in OpenSSL for it’s TLS requirements and poor optimization for heavy UDP traffic loads on Linux et al (they say it doubles CPU vs HTTP/2 for the same traffic) sounds like it’s going to be a major hurdle for widespread adoption any time soon.
Yeah a lot of the public internet is hosted by all-in-one vendors like Google Sites, Wordpress, Squarespace, etc, and object stores like S3, GCP buckets, Cloudflare, CDNs. The website owner just uploads the content, the hosting vendor does all the TLS, protocols, load balancing, whatever else. If the hosting providers update the protocols they use a huge chunk of websites will just immediately use the new one with no interaction whatsoever required on the part of the owner of the website.
If they add easy support in NGINX and HTTPD, then its easier for self-hosted endpoints to change as well, with minimal to no effort on their side.
> The book says 7% of all internet traffic already uses QUIC (HTTP/3)
The way I have understood, the book says that what is now in use (these 7%) is a "Google-only-QUIC" whereas the "standardized HTTP/3" is still used... nowhere?
The IETF QUIC remains a work in progress, perhaps to be published in 2019. HTTP/3 is an application layer on top of (IETF) QUIC, it might also be published in 2019 or later. There are implementations of current drafts, and the rough shape is settled but they're a long way from being truly set in stone and aren't in anything ordinary people use.
So unsurprisingly nobody is already doing a thing that isn't even standardised yet, but people are, as you see, writing about it.
Therefore I believe I'm right that claiming that HTTP/3 is used at all is wrong, and that 7% is not even the same QUIC that will be used with HTTP/3. So " already uses QUIC (HTTP/3)" is a wrong statement, the right can be only "GQUIC is used at the moment", also, as far as I understood, "making according to Google the 7% of the traffic" (and not on 7% of the sites as claimed). And HTTP/3 and the matching IETF QUIC are used nowhere. So, again
"> The book says 7% of all internet traffic already uses QUIC (HTTP/3)"
was wrong: the book doesn't say that, and that what is claimed that the book says (even if it doesn't) is false in more aspects.
But this book isn’t about concerning yourself with using it or implementing it, it’s about understanding what the future holds, how it works, and what roadblocks lie ahead.
The lack of API support in OpenSSL for it’s TLS requirements and poor optimization for heavy UDP traffic loads on Linux et al (they say it doubles CPU vs HTTP/2 for the same traffic) sounds like it’s going to be a major hurdle for widespread adoption any time soon.