Re: question about :authority header field

On Wed, Oct 30, 2013 at 2:57 AM, RUELLAN Herve
<Herve.Ruellan@crf.canon.fr>wrote:

> This is linked to issue #294: if we want to encourage implementations to
> omit host, we should probably leave it out of the static header table.
>
>
This is true for HTTP/2.0 client which generates HTTP/2.0 request directly.
But an intermediaries translating from HTTP/1 request to HTTP/2.0 are
instructed to omit :authority and instead use host header field for
origin-form or asterisk-form.
Since HTTP/1 does not disappear suddenly, we will continue to use host
header in HTTP/2.0 traffic in foreseeable future.

Best regards,
Tatsuhiro Tsujikawa


> Hervé.
>
> > -----Original Message-----
> > From: Martin Thomson [mailto:martin.thomson@gmail.com]
> > Sent: mercredi 23 octobre 2013 19:41
> > To: Tatsuhiro Tsujikawa
> > Cc: ietf-http-wg@w3.org
> > Subject: Re: question about :authority header field
> >
> > On 23 October 2013 05:56, Tatsuhiro Tsujikawa <tatsuhiro.t@gmail.com>
> > wrote:
> > > As described in origin-form and asterisk-form, HTTP/2 server is
> > > expected to process the request which lacks :authority as valid, where
> > > in draft-06, server rejects it if :host is missing. Is this correct?
> >
> > Your examples are all correct.
> >
> > The major change between HTTP/2.0 and HTTP/1.1 is that the host header is
> > now optional.  It can be omitted if the absolute form (i.e.,
> > :authority) is used.  In fact, we obliquely encourage implementations to
> omit
> > host.
> >
> > This places a constraint on an implementation that converts from 2.0 to
> 1.1; if
> > host is not set it has to copy it from :authority.
> >
> > But nothing has really changed other than that, host -> host and URL
> authority -
> > > :authority.
>
>