The HTTP-over-QUIC experimental protocol should be renamed to HTTP/3, officials at the (IETF) have disclosed.
There is a big gap in development fr...
For further actions, you may consider blocking this person and/or reporting abuse
Nice writeup, thanks!
Still waiting for HTTP/2 to overtake HTTP/1.1. We're still behind in adoption :(
Large, critical systems are hard to upgrade. Especially when it requires OS level changes. 'It works now, dont touch it'. Health care, defense, physical infrastructure controls, industrial control systems. These systems are critical to the safety and daily survival of millions. Upgrading the OS just to be 'up to date with the new hotness' is to much of a risk factor.
Typically these large systems only get upgraded when the hardware is replaces. If you are an American your health insurance company still interacts with AS/400 from the 1970's on a daily basis. Those machines will (nearly) never update from the version/s of software running.
Large systems update slowly, 'cause an outage could be catastrophic to the organization and public safety. The (entertainment) web can do what it wants as fast as it wants 'cause peoples lives are not at risk in daily operation.
Remember, the "internet" is every connected device. The "web" is what can be rendered in your browser.
Yeah 21 years of HTTP/1.1 do create some legacy :D
...and HTTP 0.9 was released in ... wait for it ... 1991! Nearly in step w/ HTML. Legacy is an understatement :).
Yeah, but for all intents and purposes, many developers saw gains without having to think about it much. You can't yet rely as much on some of the fancier HTTP/2 stuff, but plenty of CDNs and the like adopted it, giving immediate access to multiplexing, which should have default-helped a lot of developers.
If H/3 just makes things better when available without much implementation burden on the individual developer, should be pretty great. And H/3 seems like a huge leap.
True that!
H/2 is a leap forward and even if it's not 100% supported everywhere it's still something. It's a boon especially for this new wave of JAMstack websites. You can even stop concatenating multiple files into one because of HTTP/2.
Talking about the less fancy features: I wonder if someone could implement a HTTP/2 web server that uses machine learning onto its logs to analyze users patterns and tune server push to maximize client side performance.
Just throwing this out there :D
Isn't QUIC used over UDP?
Yes QUIC is-over-UDP.
Sorry for the confusion, I have changed the explanation. Also updated the article.
HTTP right now is on top of TCP yes. HTTP/3 will be on top of UDP. It won't matter much to web developers because HTTP already is an abstraction on TCP, we don't use TCP directly.
The only way you need to actually access the TCP/UDP layer is if you're going to write a networking server instead of just using one.
This was a great intro to whet the appetite :)
The moving picture was worth 1000+ words!
Now I am wondering how long it'd take for the world to adopt H/3. The faster we adopt, the faster we can communicate.
Anybody knows the state of FEC in QUIC? All the documents I can find conclude that it's not worth the pain... But I know use cases where it might be REALLY helpful especially if native in HTTP proxies.
This is nice for web, but seems like a insane booster for competitive gaming where latency matters so much!
Unrelated, but did you make the HTTP/3 Glitch gif you used as your cover photo?
Yes, spent more time for making the HTTP/3 Glitch gif than preparing the article.
How'd you do it? After Effects I assume?
Photoshop + online glitch effect tool.
You can get a good result using photomosh (Jitter effect) and can export the result as a .gif
This is really exciting! Looks like a huge leap from HTTP/2!
It was quic since HTTP/2
I look forward to this. TLS+H2 over UDP? Wow. What a world we live in!
It would be pretty dangerous to let HTML5 access the networking layer :-)
I'm sure in browser multiplayer games will benefit from HTTP over UDP as well.
There is WebRTC, which is, IIRC, run over UDP. If you don't care about P2P, you could conceivably use it between the server and the client, eliminating most of the signaling pain.
Very nice article.
With this http/3, the development of transmission will be much faster.
The banner image spells doomsday! :)
Great article! Thanks for sharing :)
daniel.haxx.se/http3-explained/ :D