Content-Type: text/plain;
	charset=us-ascii
Mime-Version: 1.0 (Mac OS X Mail 13.4 \(3608.80.23.2.2\))
Subject: Re: draft-ietf-quic-http-29, "1.1. Prior versions of HTTP"
From: Mark Nottingham <mnot@mnot.net>
In-Reply-To: <8c747b59-1672-4ccb-dfce-0914e74c504f@gmx.de>
Date: Fri, 19 Jun 2020 09:19:48 +1000
Cc: quic@ietf.org
Content-Transfer-Encoding: quoted-printable
Message-Id: <9B975808-6BF3-491C-8E32-C475824B0E9A@mnot.net>
References: <8c747b59-1672-4ccb-dfce-0914e74c504f@gmx.de>
To: "Julian F. Reschke" <julian.reschke@gmx.de>
Archived-At: <https://mailarchive.ietf.org/arch/msg/quic/TmD4CppstBUqF4KYuQCLMSyIQvA>
Precedence: list

My .02 -

I think we need to balance between being correct and over-complicating =
the text here. In *theory*, Julian is correct; in wide practice, =
pipelining isn't used because of the limitations it has, as well as =
implementation / interop issues.

How about something like:

"Because of limitations in how HTTP exchanges are mapped onto TCP =
connections in HTTP/1.1, multiple parallel connections are often =
used..."

?

Cheers,


> On 12 Jun 2020, at 9:48 pm, Julian Reschke <julian.reschke@gmx.de> =
wrote:
>=20
>> HTTP/1.1 [HTTP11] is a TCP mapping which uses whitespace-delimited =
text fields to convey HTTP messages. While these exchanges are =
human-readable, using whitespace for message formatting leads to parsing =
complexity and motivates tolerance of variant behavior. Because each =
connection can transfer only a single HTTP request or response at a time =
in each direction, multiple parallel TCP connections are often used, =
reducing the ability of the congestion controller to effectively manage =
traffic between endpoints.
>=20
> Isn't that a bit misleading, as at least in theory HTTP/1.1 supports
> pipelining?
>=20
> BEst regards, Julian
>=20

--
Mark Nottingham   https://www.mnot.net/