[linux-audio-dev] Gigasampler vs Halion PR war

Paul Davis paul at linuxaudiosystems.com
Tue Nov 12 10:18:01 UTC 2002


>   It isn't my intention to flame you, but has anyone here actually read and
>studied the patent? Have you? I haven't. Someone said they had a copy? Let's
>hear a listing of what the actual patent claims are before we create a
>defense.

i've read the patent. 

its innovation comes from applying a pre-existing technique
(read-ahead buffering) to a new software niche. the basic idea of "i
have this data on disk and i can't read it fast enough when i really
need to, so i should do read-ahead and cache it so that its ready to
go" is a common computer programming technique. they took that basic
idea, wrapped it up in patent language that restricts it to a specific
problem domain (playback of audio data in response to triggers), and
filed.

i don't believe that it should ever be possible to patent stuff like
this. there are vision algorithms that i think are worthy of patents -
they are truly innovative, represent new ways of approaching a
problem, and also represent deep, deep knowledge of the field. such
algorithms will not be independently reinvented by home-based linux
hackers working in their basement. the patent in question here covers
a technique that has been and will be continuously subject to
reinvention because its obvious to anyone skilled in the art. what art
is that? thats the problem. the patent office approved this probably
because it *was* genuinely novel in the domain of audio software. yet
it wasn't novel at all in the broader sense. if you approached an
averagely good programmer and said "i have this data on disk and i
can't read it fast enough when i really need to, what can i do?", they
would have instantly come up with the stuff in the patent. thats bad
practice by the patent office if you ask me. they keep the blinders on
("audio playback, audio playback, audio playback") and ignore the way
programmers actually work, which is to continually shuffle, reform and
reuse algorithms and designs across problem domains.

how many patents could be obtained by taking the design patterns book,
and filing for applications of each idea in a series of different
problem domains? say, one patent for each design as applied to

	* medical record search+display
	* audio file search+display
	* address book search+display
	* inventory search+display
	* catalog search+display

etc. etc. 

the way the patent office handles this stuff is truly and deeply broken.

--p



More information about the Linux-audio-dev mailing list