[Theora] Theora and software patents

[Ole Tange]

The list probably did not get this reply from D Richard Felker III.

Having read quite a few patents I tend to agree with Richard:  Software
patents are very broad and even if I have not read the source code of
Theora I would be surprised if it did not infringe several patents.

However, I believe the argument that On2 has not been sued for infringing
the MPEG-patents is a very strong argument. This indicates that you will
not be sued if you use Theora which basically is what people are
interested in knowing.


/Ole
-- 
The Patented Word Processor
http://ole.tange.dk/swpat/

---------- Forwarded message ----------
From: D Richard Felker III <dalias at aerifal.cx>
Date: Wed, 23 Mar 2005 17:55:35 -0500
Subject: Re: [Theora] troll or truth ?


On Tue, Mar 22, 2005 at 11:42:11PM +0100, Ole Tange wrote:
> On Tue, 22 Mar 2005, Markus Meyer wrote:
> 
> > Ole Tange schrieb:
> >
> > >So the quote: "Virtually every single patented algorithm used in mpeg*
> > >codecs is also used in theora." is correct.
> >
> > Yes, but that's about the only quote from that mail that's correct.
> 
> On Tue, 22 Mar 2005, Ralph Giles wrote:
> 
> > It is not correct. If this were the case we could not claim theora was 
> > freely implementable, and there would indeed be no point to the project.
> 
> So?! What is it? Would it be possible to construct an answer that both you
> and Richard Felker III <dalias at aerifal.cx> can agree on?
> 
> > No, as Maik stated, there are no known patent claims on theora beyond
> > the ones On2 has already granted a license on. Theora very carefully
> > does things differently from the mpeg video codecs, just as vorbis does
> > things differently from the audio codecs.
> 
> It would be interesting if Richard could point to a couple of patents that
> he believes Theora infringes - even if the patents are stupid.

I am not on the theora list and have no interest in participating in
this flamewar; however, since I was specifically CC'd I will answer
the question I was asked.

As a start, I would look at the following, which contains a list of
patents which supposedly cover mpeg2:
http://www.mpegla.com/m2/m2-essentiality.cfm

Briefly reading the Theora spec (which I will freely admit I have not
read before; I only read the specs on VP3 back when it was first
released), I expect that all patents on the following topics apply:

Forward motion compensation
Half-pel motion compensation
Motion vector precision/range
Skipped P-frame macroblocks
Macroblock prediction
IP frame followed by another P field (wtf?)
Possibly anything else relation to motion estimation or quantization.

In addition, I expect there are patents covering the loop filter used
in Theora; however these don't overlap with mpeg1/2 so they weren't
part of my original claim.

Feel free to research the above patents on your own. I will not do it
because it's not worth my time, nor worth 'tainting' myself or
whatever nonsense people think reading a patent does.

Anyway, my impression in reading the Theora spec is that it's exactly
like mpeg1/2 with B frames and all the complicated broadcast-video
features removed, loop filter added, and the whole probability
model/quantization tables in the global headers. Yes there are
significant differences in bitstream I'm sure (I'm not going to wade
through 100 pages of bitstream specs to satisfy you), but
fundamentally the idea of Theora is identical to mpeg1/2 and as such
it's covered by most (basically all, except the ones related to B
frames and broadcast features) of the same patents as mpeg1/2.

Rich