[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: MP3 decoder packaged with XMMS



Free-software-MP3 enthusiasts who believe in the protective power of
ignorance, and who are not already sticking their fingers in their
ears and shouting "la la la", may wish to do so before reading
further.  (They may also wish to start work on their affidavit
demonstrating that their local ISP prevents them from Googling "sisvel
MP3" for themselves.)

At least some MP3 player chipset vendors seem to think their customers
need patent licenses from both Sisvel (or its US subsidiary Audio
MPEG) and Thomson (sole licensor for Fraunhofer MP3 patents; see
below).  (Reportage at
http://www.electronics.globalsources.com/gsol/I/Flash-MP3/a/9000000062166.htm
; note that it misreports some Microsoft patents relating to WMA DRM
as "MP3 patents".)  Audio MPEG's website claims traction on software
implementations and (!) computers generally.  But they seem to be the
licensing agency for non-DVD-decoding-as-such use of audio patents
that Philips filed in the course of developing their DVD silicon (see
http://www.licensing.philips.com/licensees/conditions/dvd/documents1008.html
); and certain vitally interested parties don't seem to worry about
those patents when no MPEG-specific silicon is involved.

For instance, judging from the literature that came with my iPod mini
(a spiff from an employer, unsuitable for resale because of its
non-Apple-approved-warranty-voiding engraving; don't blame me for
SUPPORTING the EVIL PROPRIETARY audio DRM, mmkay?), Apple doesn't
think it needs licenses for Philips patents for its MP3 decode
functionality (implemented in software on a weird dual ARM core chip,
according to http://www.arm.com/markets/mobile_solutions/armpp/7518.html
).  They have licensed the Fraunhofer / Thomson suite, although it is
worth noting that iTunes (bundled with the iPod, and also available to
all Mac users as a free-as-in-beer download) has MP3 encoding
capabilities and Apple may have done an all-you-can-eat deal with them
covering both encoding and decoding.

Starting from the URL that Apple supplies and clicking a couple of
times, we get to http://www.iis.fraunhofer.de/amm/legal/index.html
(stating that Thomson is Fraunhofer's sole licensor) and thence to
www.mp3licensing.com.  Better not click
http://www.mp3licensing.com/patents/index.html if you don't want to
know the patent numbers involved, and you certainly had better not
(for example) Google "patent 5,924,060", click through to
patft.uspto.gov, read patentee Brandenburg's disclosure and claims,
and conclude (as I do, but IANAL and all that) that to get something
both non-obvious and actually reduced to practice by the inventor you
need to intersect at least claims 1, 3, 6, and 7.

All the patent's claims are dependent on claim 1, which is (as usual
with this sort of patent) so broad as to be obviously invalid by
itself as a description of the scope of the patent.  The claim that
reflects the principal substance of the invention is number 7.  IMHO
efficient psycho-acoustic quantization was potentially the sort of
invention suitable for patenting in 1987, depending on your
perspective on the purpose of patent law.  Granting that for the
moment, the fun is in determining which of the other claims are
essential to the invention.

If I were to consult my personal sense of what rises to a proper level
of originality, in ignorance of the details of the prior art, I would
say that 2, 4, 5, and 8 are sufficiently routine that the patent can't
be dodged by using some alternative to them.  But the meat of the
patent is in the actual reduction to practice (i. e., a concrete
quantization scheme specific to the human perception of music, of
which these claims are merely implementation details) of the
psycho-acoustic premise in claim 7, without which there is no
invention to speak of.  And until you specify both the entropy encoder
(3) and the DCT or its cousin (6), there's nothing to reduce to
practice.  IANAL, least of all a qualified US patent attorney, but
that's kind of how this first, rough-cut, "things that are obvious to
one skilled in the art" stage of the infringement analysis works as I
understand it.

YMMV; and even if the above is correct as far as it goes, a court
under the jurisdiction of the Federal Circuit might well reach
different opinions about which claims are essential to the invention
and which are not.  For one thing, there may be prior art that did
psycho-acoustic quantization using a DCT and entropy encoding (those
bits are laws of nature, and in any case already well understood in
1987) but did it less well because they chose alternatives to 2, 4, 5,
and 8.

Anyway, the alleged infringer is entitled to argue that, to the extent
that the plaintiff holds a valid patent, the court should limit its
reach (based on its claims, its disclosure, the prior art, and other
factors) in some way that causes it to fall short of covering the
allegedly infringing product.  This may or may not involve
demonstrating that this product is itself inventive to within a patent
standard, in some way that differs from an essential claim in the
claimed patent.

If I were defending, say, an Ogg/Vorbis implementation against a claim
that it infringes US Patent 5,924,060 (did I mention that IANAL?), I
would argue that a wavelet transform is sufficiently different from a
DCT (or a DFT or "a transform using Time Domain Aliasing
Cancellation", whatever that is) that Brandenburg's disclosure doesn't
teach much about how to design a psycho-acoustic quantization scheme
for wavelet coefficients.  If I were defending an MP3 decoder, I would
say instead that the decoder doesn't realize the invention because it
doesn't know or care whether the encoder used the quantization scheme
that the disclosure teaches.

Only if I were attempting to defend a literal implementation of the
disclosed method (i. e., an MP3 encoder) would I be reduced to
claiming that the entire patent is invalid because psycho-acoustic
audio encoding schemes are somehow not the sort of invention that
Congress had the power and desire to encourage when they last revised
the USPTO's charter.  I would be loth to go out on that limb, because
as far as I can see, the semi-empirical quantization thresholds in an
MP3 encoder are every bit as worthy of patent protection as the
"bottle-engaging means" in my favorite corkscrew (US Patent
#4,703,673; but see the sad story on "friction-reducing material"
claims at http://www.corkscrewnet.com/EllisCsinCourt.htm ).

In sum: I've got bad news, folks.  Arriving at an informed opinion
about where, for patent law purposes, the borderline between invention
and discovery should lie is _hard_work_.  So is the due diligence
expected of those who stake their fortune on the premise that a
presumptively valid patent should never have been granted.  You say
you want a revolution?  I suspect that, European Parliament or no
European Parliament, "No Software Patent" placards are about as
convincing to a competent court as Chairmain Mao posters were in the
Beatles' day.

Cheers,
- Michael

Disclaimer:  IANALIAJ, this is emphatically not legal advice, I
reserve the right to argue otherwise in another context, and I have
reached these opinions without applying any inside knowledge that I
may have of the legal affairs of Sisvel, Thomson, Fraunhofer, Apple,
or any other participant in the MPEG IP wars.  I did, however, make
use of the aforementioned corkscrew, strictly for demonstration
purposes.  :-)



Reply to: