W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2011

RE: part 1 section 4.1.2 - authority form

From: Eric Lawrence <ericlaw@exchange.microsoft.com>
Date: Tue, 16 Aug 2011 16:57:37 +0000
To: Daniel Stenberg <daniel@haxx.se>, Amos Jeffries <squid3@treenet.co.nz>
CC: HTTP Working Group <ietf-http-wg@w3.org>
Message-ID: <479CAD406474484E8FA0E39E694732C017D52A66@DF-M14-01.exchange.corp.microsoft.com>
FWIW, this is going to get much more relevant as WebSockets will make the practice of CONNECT'ing to non-443 ports much more common.

-----Original Message-----
From: ietf-http-wg-request@w3.org [mailto:ietf-http-wg-request@w3.org] On Behalf Of Daniel Stenberg
Sent: Tuesday, August 16, 2011 4:06 AM
To: Amos Jeffries
Cc: HTTP Working Group
Subject: Re: part 1 section 4.1.2 - authority form

On Tue, 16 Aug 2011, Amos Jeffries wrote:

> Part 2 on CONNECT appears to document this as wrong, but uses port 80 
> which is a little bit ambiguous given that CONNECT are usually used 
> for 443. Am I right in assuming that it means the port is always 
> required on CONNECT request Host: headers? (despite the obvious redundancy).

Many years ago we had curl include the port number unconditionally in Host: 
headers, only to switch it off again since there were too many proxies/servers out there that didn't like Host: headers with the default port given.

In short: when doing HTTPS (CONNECT) curl doesn't include port 443 in the
Host: header.

When doing HTTP, curl doesn't include port 80 in the Host: header.

-- 

  / daniel.haxx.se
Received on Tuesday, 16 August 2011 16:58:16 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:51:46 GMT