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

Re: MP3 decoder packaged with XMMS

On 7/18/05, Raul Miller <moth.debian@gmail.com> wrote:
> Are you suggesting that the use of time -> frequency domain mapping
> is not ostensibly covered by the presumptively valid patents?

If you want to know what I am suggesting, with regard to a particular
patent from the Fraunhofer suite (which I have looked at _very_
quickly and remember that in any case I am not qualified to judge),
read http://lists.debian.org/debian-legal/2005/07/msg00141.html .

I do not submit that I have gotten the _presumptive_ scope of the
patent straight, in terms of the rules for how you apply dependent vs.
independent claims, since in any case the patent was granted by
administrative staff who were woefully confused at the time by both
the state of the appellate law and the state of the prior art.  AIUI a
court of fact has the discretion to more or less completely rewrite
the claims of a patent when it is litigated, based on the complete
record of what was reduced to practice when by both the inventor and
other workers in the field -- although the effect is almost always to
retain the idiom of the patent's claims, striking some clauses and
merging others to form narrower independent clauses.  When I find a
reasonably "classic" and comprehensible opinion at district court
level I'll cite it.

I will say this, though: all patent agents and attorneys are expected
to be well versed in the procedures for prosecuting a patent
application through the examination system and in the administrative
standards patent examiners apply.  They generally attempt to establish
the widest possible perimeter as well as a set of more or less
defensible fall-back positions.  There may well be many patent
attorneys that also attempt to explain to their clients what is likely
to survive after a patent's claims have been evaluated by a court with
the assistance of an interested competitor, competent expert
witnesses, and a budget three to five orders of magnitude higher than
the examiner's -- but they do so under the seal of attorney-client

> Or, perhaps that all other such techniques which have been in use
> for quite some time (such as favoring frequencies which the human
> ear is sensitive to) are all not ostensibly covered by the presumptively
> valid patents?

The US number for the patent commonly cited as "the MP3 patent" is
5,579,430, which I have not yet examined closely.  But if 5,924,060 is
any indication, Karlheinz Brandenburg and his colleagues informed the
patent examiners of the existence in the prior art of
telephony-oriented audio compression techniques (such as ADPCM) which
model a waveform in the time domain and are quite poorly suited to the
signal content of recorded music and the way that the ear and brain
process it, as well as closer relatives like Zelenski's "Adaptive
Transform Coding".  They didn't claim to have invented the DCT or its
application to recorded sound.  They did claim to have reduced the
quantization of DCT coefficients to practice according to at least one
formula that is both psycho-acoustically and computationally

I don't know enough about the state of the prior art at the time of
"foreign appl. No. P3629434.9, filed Aug. 29, 1986" (the basis for
priority of 5,924,060) to say whether Herr Doktor Brandenburg was
correct in his belief that this invention met the "non-obvious" part
of the standard or that it deserved the breadth of claims stated (note
that the German patent office granted it first).  Note, in particular,
his acknowledgment that it differs from ATC principally in the use of
a single quantization level across the spectrum in a given block,
which is then iteratively adjusted to fit within a bit-rate limit. 
But to me it sounds like patentable subject matter in all the world's
major patent systems, and you'd have to fight it by focusing on the
prior art and the scope of the claims rather than shout "mathematical
method!  software patent!"

- Michael

Reply to: