correction (compression algorithm patents)
Jeffrey S. Haemer
jsh at usenix.org
Fri Jul 6 07:24:17 AEST 1990
From: jsh at usenix.org (Jeffrey S. Haemer)
Five people have now brought to my attention that my
recent editorial says the compress/uncompress algorithm is
copyrighted: Dave Grindelman, Guy Harris, Keith Bostic, Randall
Howard, and Hugh Redelmeier. That's wrong. It isn't
copyrighted, it's patented. My apologies to anyone I mislead.
Randall's note contains a lot of interesting details that it's worth posting
and he's given me permission to post it.
I've appended it.
Jeff
=====
[From Randall Howard]
Actually the problem is not that the compress algorithm is copyrighted
but that it is PATENTED by Welch (the "W" in the LZW name of the algorithm).
The patent is currently held by Unisys Corporation and they make money
from licence fees on that patent because of the use of LZW encoding
in the new high-speed modems. Note that the Lempel-Ziv algorithm
is apparently not patented, but just the Welch variant as is found in the
UNIX compress utility. Therefore, at the cost of inventing a new file
compression standard, it would be possible to escape licence fees by
using a different variant of LZ compression.
[Editor: Keith Bostic says both are patented:
original Ziv-Lempel is patent number 4,464,650,
and the more powerful LZW method is #4,558,302.
He goes on to say, however, that LZW lacks adaptive table reset
and other features in compress, and so may not apply.]
The implications of this are that no one may produce the same
output as compress produces, regardless of the program that produced
that output, without being subject to the patent. I.e., it is independent
of the actually coding used, unlike copyright. Therefore, all of the PD
versions of compress are currently in violation, as is BSD.
Representatives of Unisys at the POSIX.2 meetings claimed that
the Unisys Legal Department is pursuing the licensing of compress. In fact,
unlike copyright or trade secret protection, patent protection does not
diminish because the holder of the patent is not diligent in seeking damages
or preventing unauthorized use. Witness the large royalty payout by
Japanese semiconductor companies to Texas Instruments because they held
the patent on the concept of something as fundamental as integrated circuits.
This licence payout spans a period of over 20 years. In addition,
Unisys representatives claim that Phil Karn's PKZIP, which uses the
LZW compression algorithm, is a licenced user of the Unisys patent and
that a fee (rumoured to be somewhere in the $10,000 to $20,000 US range)
has been paid up front in lieu of individual royalties.
The ramifications for POSIX.2a are unclear. Currently, there are members
of the working group that say that they would object if a patented
algorithm were required by the standard if ANY FEE WHATSOEVER (even if $1)
were required to use it. (There are, however, precedents for standards
working in areas of patents in such areas as networking, modems, and
hardware bus structures. It appears that we software people have not
"grown up" as much when it comes to issues of licensing. Who has ever
hear of "public domain hardware"?) Some people suggested that Unisys
should allow relatively free use of the patent but should profit from
publicity rights from a citation in every POSIX.2a product manual that
contains compress. Therefore, there are currently negotiations underway
to see what kind of "special deal" Unisys would be willing to cut for the
use strictly in implementations of POSIX.2a. Depending on the outcome
of these negotiations, compress would either be dropped, re-engineered,
or retained.
Volume-Number: Volume 20, Number 101
More information about the Comp.std.unix
mailing list