[ih] QUIC history, Why did DCCP fail to gain traction?

[John Levine]

According to John R. Levine via Internet-history <johnl at iecc.com>:
>On Sun, 4 Oct 2026, Dave Crocker wrote:
>> On 10/3/2026 5:56 PM, John Levine via Internet-history wrote:
>>> HTTP version 1 and 2 ran over TCP, http/3 runs over
>>> QUIC which is built on UDP so it doesn't need kernel support, with all
>>> the multiplexing and flow control in the application
>
>> Why can't TCP run in the application?
>> What does it require that UDP+QUIC don't also require?
>
>Nothing, give or take some way to route the individual protocol 6 packets 
>to the application.  If TCP weren't already in every operating system for 
>historical reasons, I suspect we'd have a QUIC-lite for applications where 
>a single stream is adequate.

More apropos to Brian's point, if we had to wait for kernel implementations
of new protocols, we'd wait forever.  QUIC doing it backwards, first in user
code, then migrate into the kernel eventually is a lot more practical.

R's,
John
-- 
Regards,
John Levine, johnl at taugh.com, Primary Perpetrator of "The Internet for Dummies",
Please consider the environment before reading this e-mail. https://jl.ly