Re: [rtcweb] Proposal for H.263 baseline codec

Gregory Maxwell <> Fri, 20 April 2012 21:22 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id 26D9B21F8562 for <>; Fri, 20 Apr 2012 14:22:14 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -6.599
X-Spam-Status: No, score=-6.599 tagged_above=-999 required=5 tests=[BAYES_00=-2.599, RCVD_IN_DNSWL_MED=-4]
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id QwFdfe3dGl-J for <>; Fri, 20 Apr 2012 14:22:12 -0700 (PDT)
Received: from ( []) by (Postfix) with ESMTP id BD5C821F8499 for <>; Fri, 20 Apr 2012 14:22:11 -0700 (PDT)
Received: from ([]) (using TLSv1) by ([]) with SMTP ID; Fri, 20 Apr 2012 14:22:11 PDT
Received: from ([fe80::c821:7c81:f21f:8bc7]) by ([::1]) with mapi; Fri, 20 Apr 2012 14:19:12 -0700
From: Gregory Maxwell <>
To: "EXT -" <>, "" <>
Date: Fri, 20 Apr 2012 14:19:11 -0700
Thread-Topic: [rtcweb] Proposal for H.263 baseline codec
Thread-Index: Ac0fM6xo180AAEtzSLalmsQWWTSfhwAAfcYy
Message-ID: <>
References: <>,<>
In-Reply-To: <>
Accept-Language: en-US
Content-Language: en-US
acceptlanguage: en-US
Content-Type: text/plain; charset="Windows-1252"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Subject: Re: [rtcweb] Proposal for H.263 baseline codec
X-Mailman-Version: 2.1.12
Precedence: list
List-Id: Real-Time Communication in WEB-browsers working group list <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Fri, 20 Apr 2012 21:22:14 -0000 wrote:
>On 03/29/2012 09:53 AM, Stephan Wenger wrote:
>> The other issue, though (the fact that the license grant extends only to
>> the VP8 implementation as provided by google, and does not extent to
>> derivative works such as hardware implementations) should be moderately
> This is concerning, even for us open source software distributors. In
> fact, a similar situation existed some time ago with iLBC; the license
> that GIPS offered covered only the code as distributed as part of the
> RFC (although the language stating this was quite poorly constructed),

Stop. All of you are confused.

Maybe I can help—

There are _two_  VP8 patent license grants:

One for the specification:
One for the software:

These grants are subtly different in the scope of what they cover.

There are important licensing drafting reasons for a construction
which distinguishes "an implementation" and "any implementation".
In particular, the license drafter writing this kind of a royalty free
license has a tricky time making sure that the permission is broad
enough to include all the correct coverage,  but not so over-broad that
a malicious implementer could (e.g.) add a search engine to their VP8
encoder and claim that the VP8 license entitles them to search patents.

So, the drafter says something like— "those patent claims [...] that
are necessarily infringed by implementation of this specification",
which achieves the desired limited scope— and that is _all_ you get
with the MPEG-LA pool patent licenses. This is what the Google bitstream
spec license provides.

But, if we're to be serious and honest about making a royalty free format
and implementation that permission isn't quite sufficient: The reference
implementation might practice some VP8 related techniques which are not
"necessarily infringed"— they're helpful performance enhancements,
they could simply be one option out of many ways of doing the same thing,
or could just be something that was incidentally included (or even a
malicious patent trap snake in the grass)— and so it's important that
everything in the reference software also be licensed— even if its not
"necessarily infringed" by the specification.   But since it's not a
goal to also license unrelated things that a downstream user may cram in
(e.g. a search engine), that license must be limited in scope to things
which are actually in the reference implementation: "This grant does not
include claims that would be infringed only as a consequence of further
modification of this implementation." (this language is actually almost
identical to the language for this purpose in the GPL, FWIW)

So, in summary:

You are licensed, via the bitstream license, for anything Google controls
which is necessary to implement the specification— in hardware,
software, home grown, or reference implementations.  This is pretty
much as strong a grant as you can find anyone offering for any other
format specification.

You are also licensed, via the additional rights license, for any
use of the VP8 reference implementation including modified versions,
limited to the extent that the modifications don't expand the scope of
the coverage. (Exactly as the GPL explicit patent grant does)

This has already been clarified by Google on this list:
"Google confirms that the VP8 patent grant applies to both
third-party hardware and software implementations of VP8."

I admit that the fact that there are two similar but distinct licenses
is a little confusing, but I don't understand why we have to go over
it again and again on this list.  I would think that the prior Google
comment would have been sufficiently definitive.