[theora] X-Content-Duration HTTP header

Silvia Pfeiffer silviapfeiffer1 at gmail.com
Tue Dec 29 05:21:38 PST 2009

Hi ppl,

we've been using a header called X-Content-Duration on some
applications to communicate between a client and a server about the
duration of a video (or audio) file.

I've just been told by some W3C people that we should not name HTTP
headers (even if intermediate) with "X-". (see email thread excerpt

Should we move forward and register a provisional header as
"Content-Duration" and adapt all the software while there's not too
much of it out there?


---------- Forwarded message ----------
From: Julian Reschke <julian.reschke at gmx.de>
Date: Wed, Dec 30, 2009 at 12:17 AM
Subject: Re: Public feedback on HTML5 video
To: Silvia Pfeiffer <silviapfeiffer1 at gmail.com>
Cc: robert at ocallahan.org, Aryeh Gregor <Simetrical+w3c at gmail.com>,
Maciej Stachowiak <mjs at apple.com>, Gavin Carothers
<gavin at carothers.name>, HTMLwg <public-html at w3.org>,
comments at daringfireball.net

Silvia Pfeiffer wrote:
> On Mon, Dec 28, 2009 at 10:32 PM, Julian Reschke <julian.reschke at gmx.de> wrote:
>> Silvia Pfeiffer wrote:
>>> ...
>>> In Ogg and Annodex we have used a HTTP parameter called
>>> X-Content-Duration to communicate the duration of a video to a user
>>> agent. It may be an idea to introduce / implement this more widely for
>>> any content type, since typically it is not much of a problem for a
>>> server extension to calculate the duration of a video, but the user
>>> agent has to do all sorts of acrobatics to find out such. There is
>>> just the issue of creating server extensions...
>>> ...
>> *Please* do not introduce "X-*" headers. "X-" doesn't have any special
>> meaning in message headers; see
>> <http://tools.ietf.org/html/rfc3864#section-2.1> for details on registration
>> procedures.
> X-Content-Duration is used to test out such a header before putting it
> forward for registration. If ppl are happy to use such a header
> everywhere, it indeed needs to be registered properly.

But experience shows this doesn't work. Once a header gets implemented
independently, it's hard to rename it.

That's why RFC 3864 defined a "provisional" registry.

Best regards, Julian

More information about the theora mailing list