W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2013

Re: Nice

From: Martin Thomson <martin.thomson@gmail.com>
Date: Fri, 16 Aug 2013 11:12:41 -0700
Message-ID: <CABkgnnXAbZVTmwKANZaq7XgLHqfqB1bgRy+m1m_i00P1drx2qA@mail.gmail.com>
To: William Chan (陈智昌) <willchan@chromium.org>
Cc: HTTP Working Group <ietf-http-wg@w3.org>
On 16 August 2013 10:55, William Chan (陈智昌) <willchan@chromium.org> wrote:
> * Do you expect browsers to set this at all? It's not obvious to me how my
> browser would use this, so I probably wouldn't bother with it in my browser.

I wouldn't expect you to set this in your browser.  These servers are
rarely serving web pages.  Though there are cases I have in mind where
JS applications might add the header to an XmlHTTPRequest.

> * It seems like for your intended use, the priority level is not just an
> opaque level but would carry semantic meaning. If that's the case, I suggest
> renaming the header accordingly to indicate the appropriate semantic axis.

Renaming how so?  The only semantics this bears is that 0 > 1 > 2 > 3.

> * Can you give an example of where this would be used in an open,
> interoperable manner? I'd like to understand the use cases that motivate
> introducing prioritization semantics at a protocol level rather than an out
> of band mechanism (e.g. intermediaries might be coded to understand that
> different URL endpoints have certain priority levels).

I'm glad you asked.  We currently have a service deployed that
provides HTTP intermediation for devices, allowing those devices to
effectively act as HTTP servers.  Some of these devices are
battery-powered.  The interface between the intermediary and the
device is proprietary, which would allow the device to inform the
intermediary of it's current state.

If this header were present, the device - acting as a server - could
provide application-layer hints to its clients about what it might
mean to mark a request as "nice: 1" vs. "nice: 3".  Then, the
application could mark those requests that it didn't consider to be
important accordingly.

The key here is that while the intermediary to origin server
interactions are clearly proprietary, the client is just an HTTP
client in every regard.
Received on Friday, 16 August 2013 18:13:08 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:14:14 UTC