Hi.

On 2025-12-16 (Di.) 04:23, Willy Tarreau wrote:
On Mon, Dec 15, 2025 at 04:50:04PM +0100, William Lallemand wrote:
On Mon, Dec 15, 2025 at 02:24:06PM +0100, Aleksandar Lazic wrote:
Subject: Quic in "Available multiplexer protocols" list
Hi.

I have asked myself if the Output of "Available multiplexer protocols" could
be more precise.

```
Available multiplexer protocols :
(protocols marked as <default> cannot be specified using 'proto' keyword)
        quic : mode=HTTP  side=FE|BE  mux=QUIC  flags=HTX|NO_UPG|FRAMED
          h2 : mode=HTTP  side=FE|BE  mux=H2    flags=HTX|HOL_RISK|NO_UPG
          h1 : mode=HTTP  side=FE|BE  mux=H1    flags=HTX|NO_UPG
   <default> : mode=HTTP  side=FE|BE  mux=H1    flags=HTX
        fcgi : mode=HTTP  side=BE     mux=FCGI  flags=HTX|HOL_RISK|NO_UPG
        spop : mode=SPOP  side=BE     mux=SPOP  flags=HOL_RISK|NO_UPG
   <default> : mode=SPOP  side=BE     mux=SPOP  flags=HOL_RISK|NO_UPG
        none : mode=TCP   side=FE|BE  mux=PASS  flags=NO_UPG
   <default> : mode=TCP   side=FE|BE  mux=PASS  flags=
```

As far as I know is `quic` used only for `h3` for now in HAProxy right?

So maybe this could be reflected there like `quic+h3` or similar.

Opinions?


This is the list of protocols that can be used with the "proto" keyword
https://docs.haproxy.org/3.3/configuration.html#5.1-proto
So changing the name of the protocol would mean breaking other configurations,
we could have aliases thought.

In my opinion having the quic protocol in HTTP mode is quite clear for now. But
the naming could change if we have h3 over TCP, or non-HTTP over QUIC.

In fact it's a more general issue related to the fact that there are two
protocols involved (quic and h3). It's even one identified issue among
the quic working group for the work related to using QUIC multiplexing
over TLS over TCP: while TLS is part of QUIC and ALPN normally advertises
"h3", once placed on top of TLS+TCP, ALPN will have to advertise a variant
(maybe "qmux" or any such variant denoting QUIC), and h3 will have to be
advertised in some way by QUIC itself (e.g. in transport parameters maybe).

We'll need to be cautious regarding this because we can easily end up with
difficulties in configuring "server" lines in the future if we change
anything the wrong way, so for now I'd advocate for not changing anything
"just for changing".

Willy

Thank you both for the answers.

So let's wait for the outcome of the quic working group.

Regards
Aleks


Reply via email to